High Power Mode-Locked Semiconductor Lasers and Their Applications

University of Central Florida STARS

Electronic Theses and Dissertations, 2004-2019

2008

High Power Mode-locked Semiconductor Lasers And Their Applications

Shinwook Lee University of Central Florida

Part of the Electromagnetics and Photonics Commons, and the Optics Commons Find similar works at: https://stars.library.ucf.edu/etd University of Central Florida Libraries http://library.ucf.edu

This Doctoral Dissertation (Open Access) is brought to you for free and open access by STARS. It has been accepted for inclusion in Electronic Theses and Dissertations, 2004-2019 by an authorized administrator of STARS. For more information, please contact [email protected].

STARS Citation Lee, Shinwook, "High Power Mode-locked Semiconductor Lasers And Their Applications" (2008). Electronic Theses and Dissertations, 2004-2019. 3722. https://stars.library.ucf.edu/etd/3722 HIGH POWER MODE-LOCKED SEMICONDUCTOR LASERS AND THEIR APPLICATIONS

SHINWOOK LEE B.Sc. Sogang University, Republic of Korea, 1993 M.Sc. Sogang University, Republic of Korea, 1995

A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the College of Optics and Photonics/CREOL at the University of Central Florida Orlando, Florida

Spring Term 2008

Major Professor: Peter J. Delfyett, Jr.

For My Mother and Wife.

iii ABSTRACT

In this dissertation, a novel semiconductor mode-locked oscillator which is an extension of eXtreme Chirped Pulse Amplification (XCPA) is investigated. An eXtreme Chirped Pulse

Oscillator (XCPO) implemented with a Theta cavity also based on a semiconductor gain is presented for generating more than 30ns frequency-swept pulses with more than 100pJ of pulse energy and 3.6ps compressed pulses directly from the oscillator. The XCPO shows the two distinct characteristics which are the scalability of the output energy and the mode-locked spectrum with respect to repetition rate. The laser cavity design allows for low repetition rate operation <100MHz. The cavity significantly reduces nonlinear carrier dynamics, integrated self phase modulation (SPM), and fast gain recovery in a Semiconductor optical Amplifier (SOA).

Secondly, a functional device, called a Grating Coupled Surface Emitting Laser (GCSEL) is investigated. For the first time, passive and hybrid mode-locking of a GCSEL is achieved by using saturable absorption in the passive section of GCSEL. To verify the present limitation of the GCSEL for passive and hybrid mode-locking, a dispersion matched cavity is explored. In addition, a Grating Coupled surface emitting Semiconductor Optical Amplifier (GCSOA) is also investigated to achieve high energy pulse. An energy extraction experiment for GCSOA using stretched pulses generated from the colliding pulse semiconductor mode-locked laser via a chirped fiber bragg grating, which exploits the XCPA advantages is also demonstrated.

Finally, passive optical cavity amplification using an enhancement cavity is presented. In order to achieve the interferometric stability, the Hänsch-Couillaud Method is employed to stabilize the passive optical cavity. The astigmatism-free optical cavity employing an acousto-

iv optic modulator (AOM) is designed and demonstrated. In the passive optical cavity, a 7.2 of amplification factor is achieved with a 50 KHz dumping rate.

v ACKNOWLEDGMENTS

First of all, I would like to thank to my adviser, Prof. Delfyett, who has introduced the world of ultrafast optics. Personally and scientifically, I have learned a lot of things from him. His positive attitude and strong enthusiasm for life is also what I would like to earn during my Ph.D course.

I have enjoyed a supportive atmosphere of the ultrafast group in CREOL. It was fortunate for me to have such supportive and good colleagues. They are Prof. Sangyun Gee, Dr. Michael Mielke,

Dr. Tolga Yilmaz, Dr. Myoung Taek Choi, Dr. Kyungbum Kim, Dr. Wangkuen Lee, Dr. Luis

Archundia, Dr. Bojan Resan, Dr. Nishant Bhatambrekar, Franklyn Quinlan, Sarper Ozharar, Ji-

Myoung Kim, Mr. Dimitrios Mandridis, Ibrahim Ozdur, Mohammad Umar Piracha, Charles

Williams, Iffat Nayyar, Sharad Bhooplapur, Abhijeet Ardey, Henock Legesse, Dat Nguyen,

Nazanin Hoghooghi, Josue Davila-Rodriguez, Sen-Yong Chen and my lab mate, Dimitrios

Mandridis.

I express my sincere thanks to all of professors in CREOL and staffs for establishment of such a solid structure of photonics research. Especially, I also want to acknowledge my committee members for this dissertation, who are Prof. Likamwa, Prof. Riza, and Prof. Kapoor.

Finally, I dedicate this dissertation to my wife, Kyunglae Kang, and my mother, Chiae Kim.

Without their prayer and love, this dissertation could not have been done.

vi TABLE OF CONTENTS

ABSTRACT...... iv

ACKNOWLEDGMENTS ...... vi

TABLE OF CONTENTS...... vii

LIST OF FIGURES ...... xi

LIST OF TABLES...... xviii

LIST OF ACRONYMS/ABBREVIATIONS...... xix

CHAPTER 1 INTRODUCTION ...... 1

1.1 Semiconductor gain media ...... 1

1.1.1 Semiconductor laser...... 1

1.1.2 Semiconductor optical amplifiers ...... 4

1.2 Carrier dynamics of semiconductor gain media...... 7

1.3 Semiconductor mode-locked lasers...... 11

1.3.1 Active mode-locking...... 13

1.3.2 Passive and hybrid mode-locking ...... 14

1.4 Scope and organization of the dissertation...... 17

CHAPTER 2 EXTREME CHIRPED PULSE AMPLIFICATION...... 20

2.1 High power semiconductor mode-locked laser ...... 20

2.2 Concept of extreme chirped pulse amplification...... 21

2.3 Colliding pulse semiconductor mode-locked laser...... 23

2.4 Chirped fiber bragg grating for XCPA...... 32

2.5 Extreme chirped pulse amplification system...... 34

vii 2.6 Summary...... 37

CHAPTER 3 EXTREME CHIRPED PULSE OSCILLATOR BASED ON THETA CAVITY

DESIGN...... 39

3.1 A more of extreme chirped pulse amplification ...... 39

3.2 Preliminary experiment for extreme chirped pulse oscillator ...... 41

3.3 Theta cavity design for extreme chirped pulse oscillator...... 45

3.4 Extreme chirped pulse oscillator for > 1ns stretched pulses ...... 55

3.4.1 Theta cavity and experimental setup...... 56

3.4.2 Experimental results...... 62

3.5 Amplification of stretched pulses from the XCPO...... 69

3.6 <100MHz repetition rate operation of the Theta cavity ...... 74

3.7 Preliminary experiment of feed forward intensity noise cancellation and gain flattening

scheme for frequency swept pulses ...... 78

3.8 Summary...... 80

CHAPTER 4 GRATING COUPLED SURFACE EMITTING LASER AND GRATING

COUPLED SURFACE EMITTING SEMICONDUCTOR OPTICAL AMPLIFIER ...... 82

4.1 Introduction ...... 82

4.2 Grating coupled surface emitting laser...... 85

4.3 Passive and hybrid mode-locking of grating coupled surface emitting laser...... 89

4.3.1 Experimental setup for mode-locking...... 89

4.3.2 Passive mode-locking of GCSEL...... 91

4.3.3 Hybrid mode-locking of GCSEL ...... 94

4.3.4 Hybrid mode-locking with dispersion matched cavity of GCSEL ...... 96

viii 4.4 Grating coupled surface emitting semiconductor optical amplifier ...... 100

4.4.1 Grating coupled surface emitting semiconductor optical amplifier...... 100

4.4.2 GCSOA and CW amplification...... 101

4.4.3 GCSOA and extreme chirped pulse amplification for energy extraction ...... 103

4.5 Summary...... 109

CHAPTER 5 PASSIVE OPTICAL CAVITY AMPLIFICATION...... 111

5.1 Introduction of passive optical cavity...... 111

5.2 Hänsch-Couillaud stabilization...... 112

5.3 Experimental setup and results...... 114

5.3 Summary...... 124

CHAPTER 6 SUMMARY...... 125

6.1 Finished work ...... 125

6.2 Future direction ...... 126

APPENDIX A: TRANSFORM-LIMITED PULSE CALCULATION...... 130

APPENDIX B: DESIGN CDEO FOR PASSIVE OPTICAL CAVITY USING ASAPTM ...... 133

APPENDIX C: EXTERNAL CAVITY OF 20SECTION SEMICONDCUTOR MODE-

LOCKED LASER...... 139

APPENDIX D: NOVEL FREQUENCY SWEPT SOURCE OF MODE-LOCKED

SEMICONDUCTOR LASER ...... 142

LIST OF REFERENCES...... 145

CHAPTER 1...... 145

CHAPTER 2...... 147

CHAPTER 3...... 148

ix CHAPTER 4...... 150

CHAPTER 5...... 152

CHAPTER 6...... 153

x LIST OF FIGURES

Fig. 1 - 1. Semiconductor lasers (a) Schematic of semiconductor laser, and (b) picture of

semiconductor lasers with C-mounts [8]...... 2

Fig. 1 - 2. Energy-wave number (E-k) diagram for semiconductor gain medium...... 3

Fig. 1 - 3. Effect of residual reflection from facets of SOAs. The schematics and the gain ripples

are exaggerated for clarity...... 6

Fig. 1 - 4. Qualitative illustration of the evolution of the carrier distribution (electron density ρ vs.

electron energy E) in the SOAs [14]...... 9

Fig. 1 - 5. Short pulse amplifications of (a) long pulse regime (>2ps) and (b) short pulse regime

(<0.5ps)...... 11

Fig. 1 - 6. Mode-locked pulse train...... 12

Fig. 1 - 7. Active mode-locking...... 14

Fig. 1 - 8. Illustration of transmission and pulse shaping of saturable absorber...... 15

Fig. 1 - 9. Passive mode-locking of gain and loss dynamics...... 16

Fig. 2 - 1. Schematic of mater oscillator and power amplifier (MOPA) system. Semi-MLL:

semiconductor mode-locked laser...... 21

Fig. 2 - 2. The eXtreme Chirped Pulse Amplification (XCPA). Semi-MLL: semiconductor mode-

locked laser, tp: stretched pulse duration...... 22

Fig. 2 - 3. Saturable Absorber(a) The schematics of Saturable Absorber [8], (b) exitonic

absorption at 970nm...... 25

xi Fig. 2 - 4. Colliding pulse semiconductor mode-locked lasers. (a) Bi-directional ring cavity, (b)

anti-resonant linear cavity, (c) linear cavity, and (d) unidirectional ring cavity...... 28

Fig. 2 - 5. Picture of colliding pulse semiconductor mode-locked laser...... 29

Fig. 2 - 6. The CPSML’s performance. (a) Sampling scope trace, (b) mode-locked spectrum, and

Fig. 2 - 7. Chirped fiber bragg grating (a) schematic, (b) group delay...... 33

Fig. 2 - 8. eXtreme Chirped Pulse Amplification Setup [5]...... 35

Fig. 2 - 9. The XCPA performance [5]. (a) Spectrally resolved streak camera trace of the

stretched pulse, (b) compressed autocorrelation...... 37

Fig. 3 - 1. Preliminary setup for eXtreme Chirped Pulse Oscillator...... 42

Fig. 3 - 2. Performance of oscillator. (a) Sampling scope trace of compressed pulse, and (b)

sampling scope trace of stretched pulse...... 43

Fig. 3 - 3. Performance of oscillator. (a) Intensity auto-correlation, and (b) mode-locked

bandwidth...... 44

Fig. 3 - 4. Experimental configuration for the extreme chirped pulse mode-locked diode laser

ring oscillator...... 46

Fig. 3 - 5. Relative reflectances, transmittances and group delays of the two ports of the CFBG.

...... 47

Fig. 3 - 6. Temporal response of an electric comb generator...... 48

Fig. 3 - 7. Sampling scope image of compressed pulse at 1.95GHz...... 49

Fig. 3 - 8. Autocorrelation of 200mA, 400mA, 600mA and transform-limited pulse at 1.95GHz.

...... 50

xii Fig. 3 - 9. Externally compressed auto-correlation of the compressed pulse...... 51

Fig. 3 - 10. Time-resolved chirp measurement and sampling scope image of stretched pulse. ... 52

Fig. 3 - 11. Mode-locked spectrum of stretched pulse...... 53

Fig. 3 - 12. L-I curve for stretched pulse and compressed pulse output ports with 10%

transmission output couplers, respectively...... 54

Fig. 3 - 13. Performance of the 2000ps/nm CFBG. (a) Group delay, (b) insertion loss, (c) group

delay ripple, and (d) sum of group delay ripple...... 58

Fig. 3 - 14. Electrical comb generator at 100MHz...... 59

Fig. 3 - 15. Diagram Theta cavity for Extreme Chirped Pulse Oscillator (XCPO) made with an

SOA...... 60

Fig. 3 - 16. Enclosed box for eXtreme Chirped Pulse Oscillator...... 61

Fig. 3 - 17. The stretched pulse performance of the Theta cavity. (a) Spectrally resolved streak

camera image, (b) mode-locked spectrum at 505MHz, (b) 246MHz, and (c) 102MHz. All

data are taken at 600 mA DC-biased SOA...... 63

Fig. 3 - 18. Scalability of pulse energy and spectrum. (a) Mode-locked spectral bandwidth vs.

repetition period, and (b) Energy per pulse for 505MHz, 246MHz, and 102MHz repetition

rate...... 64

Fig. 3 - 19. Sampling scope traces at (a) 246MHz, and (b) 102MHz...... 65

Fig. 3 - 20. Second harmonic intensity autocorrelation at 102MHz. Inset is the mode-locked

spectrum for generating short pulses...... 66

Fig. 3 - 21. 660ps/nm CFBG and short pulse. (a) Group delay and Reflection, (b) sampling

scope trace...... 67

xiii Fig. 3 - 22. eXtreme Chirped Pulse Oscillator with 660ps/nm CFBG. (a) Spectrally resolved

streak camera of the stretched pulse and (b) mode-locked spectrum...... 68

Fig. 3 - 23. Slab-Coupled Optical Waveguide Amplifier. (a) Measured gain, and (b) SCOWA

package...... 70

Fig. 3 - 24. SCOWA characteristics. (a) Amplified spontaneous emission, and (b) I-V curve.... 70

Fig. 3 - 25. SCOWA amplification. (a) Spectrally resolved steak camera trace, and (b) increased

power vs. current...... 71

Fig. 3 - 26. Amplified spectrum. (a) The stretched pulse spectrum of the Theta cavity, and (b)

after SCOWA amplification...... 72

Fig. 3 - 27. RF spectrum of photo-detected signal...... 74

Fig. 3 - 28. 33MHz repetition rate operation of the Theta cavity with 660ps/nm CFBG. (a) Short

pulse on sampling scope, and (b) frequency-swept pulse on spectrally resolved streak camera.

...... 75

Fig. 3 - 29. Group delay ripple (GDR) of CFBG and mode-locked spectrum...... 76

Fig. 3 - 30. Feed forward scheme for frequency swept pulses. (a) General schematic, and (b)

concept...... 78

Fig. 3 - 31. Feed forward noise cancellation scheme. (a) Before feed forward scheme, (b) after

feed forward scheme, and (c) experimental setup...... 79

Fig. 4 - 1. Grating-Coupled Surface-Emitting Laser...... 83

Fig. 4 - 2. A scanning electron microscope (SEM) image of grating...... 86

Fig. 4 - 3. The GCSEL performance (a) Light vs Current curve of grating port, (b) Spontaneous

emission spectrum of cleaved facet, (c) Spontaneous emission spectrum of grating port. .... 88

xiv Fig. 4 - 4. The experimental setup for passively and hybridly mode-locked external cavity using

GCSEL and sketch of GCSEL...... 90

Fig. 4 - 5. (a) Autocorrelation trace of compressed and uncompressed optical pulses in passive

mode-locking. (b) Hysteresis of passive mode-locking and LI curve of cavity (inset). (c) Near

field profile before passive mode-locking. (d) Near field profile...... 93

Fig. 4 - 6. (a) Fundamental frequency RF spectrum measured with 100KHz resolution

bandwidth in hybrid mode-locking and passive mode-locking (inset). (b) Optical spectrum in

hybrid mode-locking. (c) Autocorrelation trace of compressed optical pulses in hybrid mode.

...... 95

Fig. 4 - 7. Dispersion matched cavity of GCSEL. (a) Experimental schematic, and (b) optical

layout with respect to vertical and horizontal directions...... 97

Fig. 4 - 8. Hybrid mode-locking performance. (a) Intensity auto-correlation, and (b) its mode-

locked spectrum...... 98

Fig. 4 - 9. Proposed 2-section GCSEL. (a) Lateral current injection laser diode [8], and (b) 2-

section GCSEL based on current wafer structure...... 99

Fig. 4 - 10. Typical scheme of Grating-Coupled Surface-Emitting Semiconductor Optical

Amplifier...... 101

Fig. 4 - 11. CW injection to GCSOA driven by QCW (a) experimental setup, (b) amplified peak-

power vs. input power, (c) spectrum of the oscillator, and (d) injected spectrum vs. amplified

spectrum...... 102

Fig. 4 - 12. Energy extraction from GCSOA setup and general performance by using XCPA. (a)

experimental setup, (b) spectrum of CPSML, (c) autocorrelation of CPSML , and (d)

amplified spectrum of GCSOA...... 104

xv Fig. 4 - 13. Mode-matched optical system (a) setup, (b) output spatial mode before injection and

Fig. 4 - 14. XCPA and GCSOA performance. (a) Peak powers for 20nsec stretched pulse, (b)

spectral evolution as increasing biased current, and (c) optical signal to noise ratio...... 108

Fig. 5 - 1. Passive optical cavity. (a) Time domain [1], and (b) frequency domain explanation.

...... 112

Fig. 5 - 2. Passive optical cavity schematic...... 116

Fig. 5 - 3. Picture of passive optical cavity...... 117

Fig. 5 - 4. Passive optical cavity (POC) design and beam profiles. The design beam profiles at 2

different positions (near AOM) of and a measured beam profile at one position...... 118

Fig. 5 - 5. The Fabry-Perot resonance, the error signal and the scanning voltage of the POC with

scanning Piezo-electric actuator...... 119

Fig. 5 - 6. The dumping electronics before a high power amplifier...... 120

Fig. 5 - 7. Dumped energy from the POC with respect to the dumping rate and dumping interval.

...... 121

Fig. 5 - 8. Photon life time of the POC...... 122

Fig. 5 - 9. Autocorrelations (a) from the passive optical cavity dumping and (b) from the

oscillator...... 123

Fig. 6 - 1. The P-ADC configurations. (a) TDM [10], (b) WDM [10], and (c) time stretch system

[11]...... 128

Fig. A - 1. Diagram for transform-limited pulse and autocorrelation...... 131

xvi Fig. C - 1. 2-Section external cavity semiconductor mode-locked laser...... 140

Fig. C - 2. Performance of 2-Section external cavity semiconductor mode-locked laser...... 141

Fig. D - 1. Proposed frequency swept semiconductor mode-locked laser...... 143

Fig. D - 2. Mode-locked spectrum with fixed repetition rates...... 144

xvii LIST OF TABLES

Table 1 - 1. Main features of fiber amplifiers and SOAs [9]...... 5

Table 2 - 1. QW saturable absorber wafer structure...... 24

Table 3 - 1. Wall-Plug Efficiency (WPE) of the Theta cavity and a SCOWA amplification

system...... 73

Table 4 - 1. GCSEL wafer structure...... 85

Table 4 - 2. The parameters for GCSEL and GCSOA which are used in this experiment...... 106

Table 5 - 1. Comparison between Hänsch-Couillaud method and PDH method [10]...... 114

xviii LIST OF ACRONYMS/ABBREVIATIONS

AOM Acousto-Optic Modulator

AR Anti-Reflection

CCD Charge Coupled Device

CFBG Chirped Fiber Bragg Grating

COMD Catastrophic Optical Mirror Damage

CPA Chirped Pulse Amplification

CPSML Colliding Pulse Semiconductor Mode-locked Laser

CW Continuous Wave dB Decibels

E-ADC Electronic Analog Digital Conversion

Ed Erbium

EDFA Erbium Doped Fiber Amplifier

EUV Extreme Ultra-Violet

GCSEL Grating Coupled Surface Emitting Laser

GCSOA Grating Coupled surface emitting Semiconductor

Optical Amplifier

GDR Group Delay Ripple

HWP Half Wave Plate

LOC Large Optical Cavity

MOPA Master Oscillator Power Amplifier

Nd Neodymium

xix OCT Optical Coherence Tomography

P-ADC Photonic Analog Digital Conversion

PBS Polarization Beam Splitter

PDH Pound-Drever-Hall

PID Proportional Integral Derivative

POC Passive Optical Cavity

QCW Quasi Continuous Wave

QWP Quarter Wave Plate

SCOWA Slab Coupled Optical Waveguide Amplifier

SEM Scanning Electron Microscope

Semi-MLL Semiconductor Mode-locked Laser

SOA Semiconductor Optical Amplifier

SPM Self Phase Modulation

TDM Time Division Multiplexing

TEC Thermal Electric Cooler

TRC Time Resolved Chirp

VECSEL Vertical External Cavity Surface Emitting Laser

WDM Wavelength Division Multiplexing

WPE Wall Plug Efficiency

XCPA eXtreme Chirped Pulse Amplification

XCPO eXtreme Chirped Pulse Oscillator

XUV eXtreme Ultra-Violet

Yd Ytterbium

xx CHAPTER 1

INTRODUCTION

1.1 Semiconductor gain media

1.1.1 Semiconductor laser

Since the demonstration of the first semiconductor laser [1] in 1962, semiconductor laser diodes have received a great deal of attention for research and industrial applications. Typical dimensions of the semiconductor laser diodes are less than 10mm long and 1~2 mm wide. The semiconductor laser diode is simply a semiconductor gain medium with reflective mirrors.

The superior advantages of semiconductor lasers such as a compactness, excellent wall- plug efficiency, wavelength agility, mass production capability and simple pumping mechanism are attractive points for many applications. For the past few years, many novel designs have been developed to increase power extracted from semiconductor lasers using different structures.

Inverse bow-tie laser diodes [2], tapered structure laser diodes [3], flared structure laser diodes

[4], and slab coupled optical wave guide laser diodes [5] have been invented and evolved. Even surface emitting laser diode structures [6, 7] have been demonstrated with keen attention from many researchers due to their unique properties such as a single mode beam profile and the possibility to reduce the catastrophic optical mirror damage (COMD), which is one of the practical limitations for semiconductor lasers. The basic laser diode picture with C-mounts and the schematic are shown in Fig. 1-1.

(a) (b)

Fig. 1 - 1. Semiconductor lasers (a) Schematic of semiconductor laser, and (b) picture of

semiconductor lasers with C-mounts [8].

The gain, and hence the optical power of the laser diode is due to the inter-band transition

between conduction band and valence band which can be brought about by electrical pumping or optical pumping from external sources. Proper energy of pumping sources excites electrons from the valence band to the conduction band and the electrons quickly settle down to the bottom of the valence band. At the same time holes are generated and are moved to the top of the conduction band. Light can be emitted by the recombination of electron-hole pairs through spontaneous or stimulated emission with injection of photons having the corresponding band gap energy. Hence the laser diode can be simply modeled by the three-level energy system. If there are reflections or optical feedback from the facets of semiconductor gain medium and the reflection is amplified, one can build laser due to population inversion. The general idea of the

light emitting from the semiconductor gain media is depicted in Fig. 1-2. Due to the energy

2 equation, the E-k relation is a simple parabola for a conduction band and a valence band.

However, their eccentricities are different because of differences of mass of the electron and hole.

Fig. 1 - 2. Energy-wave number (E-k) diagram for semiconductor gain medium.

As mention above, one of the big advantages of the semiconductor laser diode is wavelength agility due to the different material composition from the semiconductor. From

400nm (GaN) to 10um (Quantum Cascade Laser), semiconductor lasers can emit a broad range of wavelengths. Some of the materials used for making semiconductor laser are listed below.

GaAs (gallium arsenide)

AlGaAs (aluminum gallium arsenide)

GaP (gallium phosphide)

InGaP (indium gallium phosphide)

GaN (gallium nitride)

InGaAs (indium gallium arsenide)

3 GaInNAs (indium gallium arsenide nitride)

InP (indium phosphide)

GaInP (gallium indium phosphide)

1.1.2 Semiconductor optical amplifiers

A semiconductor optical amplifier (SOA) is simply a laser diode without optical feedback.

In order to suppress the optical feedback, anti-reflection coatings on both sides of semiconductor gain medium are deposited. Hence SOAs can be used to amplify an input signal due to gain.

Since optical communication or high power laser system require amplification of an input signal, optical amplifier technology has rapidly developed. Rare earth materials doped fiber amplifiers are also widely used to increase the optical signal. Hence, the optical amplifiers can be divided into two classes [9]. One is fiber amplifiers with rare earth materials such as Ed, Yb or

Nd doped and the other is SOAs. Each has advantages and disadvantages depending on the application. Typical differences between them are compared on Table 1-1.

Table 1 - 1. Main features of fiber amplifiers and SOAs [9].

Feature Fiber AmplifierSOA

Typical maximum internal gain (dB) 30-50 30

Typical insertion loss (dB) 0.1-2 5-10

Polarization sensitive? No Yes

Pump source Optical Electrical

3dB gain bandwidth 30 30-50

Nonlinear effects Negligible Yes

Saturation output power (dBm) 10-15 5-20

Typical intrinsic noise figure (dB) 3-5 5-7 Photonic integrated circuit No Yes compatible? Function device possibility? No Yes

In order to amplify a signal without adding additional noise such as Fabry-Peron etalon effect, a traveling wave operation is needed. Or in special cases, such as, mode-locking, very stringent AR coating performance is needed to suppress resonance effects from SOA. Usually less than 10-5 reflectivity is required to maintain the high quality mode-locking from SOAs.

Tilting waveguides to make angled facet SOAs [10] or in situ monitoring method during the AR coating deposition [11] are used to achieve less than 10-5 reflectivity [12]. The residual reflectivity due to incompleteness of AR coating degrades the performance of SOAs. One of

5 problems is gain ripples due to feed back reflections from the facets as shown in Fig. 1-3. This gain ripple is defined by cavity longitudinal mode spacing.

c Δ=ν , [1-1] 2Lneff where, Δν is the longitudinal mode spacing, c is the velocity of light, L is the length of a device,

neff is the effective refractive index.

Fig. 1 - 3. Effect of residual reflection from facets of SOAs. The schematics and the gain ripples

are exaggerated for clarity.

The modulation frequency in the Fig. 1-3 corresponds to the longitudinal mode spacing when residual reflection occurring.

6 1.2 Carrier dynamics of semiconductor gain media

In this section, we briefly review the carrier dynamics of SOAs to understand further

chapters.

The simplified energy model of an SOA is the conduction and valence band model shown in Fig. 1-2. The typical time evolution of carrier dynamics can be expressed by using the following equation for indirect band gap semiconductors.

∂N =−R RR − − R ∂t inj rad nrad st , [1-2]

where N is the total carrier density, Rinj is the carrier injection rate by the injection current or

optical pumping, Rrad is the radiative recombination rate of spontaneous emission in the active

region, Rnrad is the non-radiative recombination rate, and Rst denotes stimulated recombination rate, respectively.

Among the non-radiative recombination, Auger recombination is usually the most important recombination mechanism. There are many forms of Auger recombination process [9] and other sources for non-radiative recombination. Material defects and surface effects cause significant non-radiative recombination. Material defects in the active regions can occur during device fabrication and increase as the device ages. End facets of SOAs absorb impurities from the air that can be a source of non-radiative recombination. This non-radiative recombination represents the trap non-radiative recombination. Another non-radiative recombination source is carrier leakage. This effect arises due to drift or diffusion of carriers across the SOA heterojuctions. Hence we can summarize the non-radiative recombination as following.

7 Rnrad=+RR tr aug + R leak , [1-3]

where Rtr is the trap non-radiative recombination, Raug is the Auger recombination, and

Rleak represents leak non-radiative recombination.

When considering carrier life time, the radiative recombination and the non-radiative recombination are taken into account as following.

N τ s = , [1-4] RRrad+ nrad

where τ s denotes the carrier life time. Usually the carrier life time of SOAs with quantum well is from a few hundred picosecond to nanosecond.

Another thing to be considered is the gain to determine how much input signal can be

amplified in SOA. The material gain of semiconductor gm can be written as following.

gaNNm = ()− 0 , [1-5]

where a is the differential gain with respect to N , and N0 is the transparency carrier density.

This equation is valid for long pulse amplification (>10ps) or weak input signal amplification.

However, this equation is a little bit modified at short pulse amplification (<10ps) or when the

input power is too high. The gain equation at this case is expressed as g g = m 1+ ε S , [1-6]

where g is the gain, S is the photon density, and ε denotes the non-linear gain suppression

factor. Non-linear gain suppression in SOAs is mainly due to carrier heating and spectral hole

burning [13].

8 As SOAs are used to amplify modulated signals or pulses, it is of interest to investigate

the dynamic behavior of carrier. When an input signal is injected into an SOA, the excited electrons in the valence band combine with the holes in the conduction band to cause stimulated

emission leads to amplification in the SOA. If we use a pulse as an input signal to monitor the

carrier dynamics of SOAs, we can illustrate the carrier dynamics as a qualitative evolution as

shown Fig. 1-4.

Fig. 1 - 4. Qualitative illustration of the evolution of the carrier distribution (electron density ρ vs.

electron energy E) in the SOAs [14].

9 Within a few picoseconds, a quasi-equilibrium distribution at the lattice temperature is

established. Before reaching the steady state of Fermi-Dirac distribution, the carrier distribution

in the non-equilibrium state is mainly governed by spectral hole burning, in order words, by a

localized reduction in the number of carriers at the transition energies and by carrier heating. It

describes the fact that the carrier temperature is different from the lattice temperature. There are

two different times related with carrier heating. The first is the carrier-photon scattering time,

which is the average time carriers require to cool down to the lattice temperature. The second is

the carrier-carrier scattering time, which is the average time taken by the carrier population to

reach a heated equilibrium from the initial non-heated equilibrium. Both spectral hole burning and carrier heating have characteristic times on the order of hundreds of femtoseconds. It is well known that the carrier heating and spectral hole burning cause gain suppression and bandwidth

suppression in semiconductor lasers [14]. Thus carrier dynamics can be characterized with pump-probe measurements which employ femtosecond optical pulses.

We consider another aspect of amplification in SOAs in terms of pulses. When a short pulse is passing through an SOA, the gain is depleted, which leads to a change in the refractive index. The change of refractive index is inversely proportional to the gain. Consequently, this refractive index change over the pulse duration results in an instantaneous frequency due to self phase modulation (SPM). This strong carrier dynamics in SOAs lead to pulse broadening in the

temporal domain and spectral broadening in the spectral domain [15, 16], especially for short

pulses, on the order of 1ps. In cases when the input pulse is < 1ps, the carrier heating and

cooling effects can be observed. The basic amplification dynamics of SOA as it relates to the

input pulse width is summarized in Fig. 1-5.

gain

d n gain

d n

Carrier freq. ω Carrier 0 freq. ω 0

inst. freq. -5 -4 -3 -2 -1 0 1 2 3 4 5 inst. freq. -3 -2 -1 0 1 2 3

Time (ps) Time (ps) (a) (b)

Fig. 1 - 5. Short pulse amplifications of (a) long pulse regime (>2ps) and (b) short pulse regime

(<0.5ps).

1.3 Semiconductor mode-locked lasers

The laser emission often consists of a set of resonant mode ωm = 1, 2, 3,…., separated by

δω . These modes are called longitudinal modes of laser cavity. We can simply express electric

field in the cavity by following,

Et()=+⋅+∑ Amm exp[( jω0 mδω ) t φ ] [1-7] m

11 where Am and φm represent the amplitude and phase of the m -th mode. If there is no fixed

relationship in phase, which is often the general case, then the laser output varies randomly.

However, the phase and amplitude can be forced to maintain a fixed value, for example,

φmm−=φδφ−1 [1-8]

Ajmmexp (φ )=+⋅ Ajm00 exp (φδφ ) [1-9]

The output of the laser cavity is, then, a periodic function in time. This phenomenon is called

mode-locking. Fig. 1-6 shows a superposition of 10 coherent sinusoidal waves. The advantages of mode-locking over gain-switching and Q-switching are high qualities of optical combs and temporal pulses. The semiconductor gain medium is widely used for mode-locking to generate short pulse due to fast life time which is good for high speed network, and rigidness in terms of environmental influences.

Fig. 1 - 6. Mode-locked pulse train.

Mode-locking can be divided by 3 categories which are active mode-locking, passive mode-locking, and hybrid mode-locking. (a) Active mode-locking is achieved by external gain or loss modulation. In semiconductor gain media, a modulator or direct gain modulation is used to generate short pulse in active mode-locking. (b) Passive mode-locking takes advantages of nonlinear medium having saturable absorption. Hence no external modulation is required to achieve passive mode-locking. (c) Hybrid mode-locking is the combination of active mode-locking and passive mode-locking to overcome drawbacks of each mode-locking. Therefore it uses a storable absorber and external modulation.

1.3.1 Active mode-locking

Active mode-locking is achieved with external modulation of gain or loss in the cavity at a frequency that is the cavity fundamental or integer multiples of the cavity fundamental. In time domain analysis, as maximum gain matches the modulation peak, short optical pulses are generated. Loss or gain modulation over threshold generates a very short window of net gain in the laser. In frequency domain analysis, this modulation generates side bands which can be made to match the cavity resonance. This can result in the enhancement of the optical combs of the cavity with a fixed phase relation ship. Active mode-locking of semiconductor laser is schematically depicted in Fig 1-7.

Fig. 1 - 7. Active mode-locking.

1.3.2 Passive and hybrid mode-locking

Passive mode-locking uses a saturable absorber to generate short pulses. Without any external modulation, time-dependent gain and loss dynamics enable us to generate short pulses.

The schematic mechanism is depicted in Fig. 1-8. The nonlinear transmission from a saturable absorber is due to its bleaching effect so that if a short pulse passes though a saturable absorber, the leading edge of a pulse is lower than that of the trailing edge.

14 One of requirement for passive mode-locked lasers is that the saturation of loss should be

faster than that of gain, which implies the saturation intensity of loss should be less than that of

saturation intensity of the gain as shown in following equation [17], ω AAω EEag≡<≡hh s ∂∂aN ∂∂ gN s [1-10]

a where Es is the saturation energy of the absorber, A is the cross section area, ∂∂aN is the

g differential absorption, ∂∂gNis the differential gain, and Es is the saturation energy of the gain.

As shown above equation, sometimes the cross-sectional area of the saturable absorber is the key

parameter to achieve high quality passive mode-locking.

Fig. 1 - 8. Illustration of transmission and pulse shaping of saturable absorber.

When a laser is turned on, the loss of the cavity is usually larger than gain. If the intracavity power increases, then the saturable absorber experiences the saturable absorption or carrier bleaching. Hence the trailing edge of the pulse experiences a “transparent” saturable

15 absorber while the leading edge of the pulse experiences the absorption. Proper conditions enable the pulses to shorten and the spectrum to widen after successive round trips in the cavity.

Between pulses, spontaneous emission is suppressed due to the absorption in the saturable absorber.

Passive mod-locking requires that the recovery time of the saturable absorber is faster than that of the semiconductor gain medium and saturable gain is greater than the saturable loss in the cavity. The schematic drawing of passive mode-locking is shown in Fig. 1-9 [17].

Fig. 1 - 9. Passive mode-locking of gain and loss dynamics.

16 Due to the nonlinear absorption inside the semiconductor saturable absorber, if we

operate laser in CW regime, we can observe hysteresis between the downward current injection

and the upward current injection which is a good indication of the bleaching process of carriers.

Hybrid mode-locking is the combination of active mode-locking and passive mode-

locking to combine the advantages of both. One of disadvantages of active mode-locking is very

difficult to achieve ultra short pulse while that of passive mode-locking is that the detected photon signal RF phase noise is higher than that of active noise because there is no reference electrical driving force. In order to overcome the disadvantages of both, an external RF signal is injected into the laser cavity using gain modulation or loss modulation in conjunction with a saturable absorber. In semiconductor mode-locked laser, the proper position of semiconductor gain media and the saturable absorber is also important in hybrid mode-locking due to matching electrical signal and optical pulses.

1.4 Scope and organization of the dissertation

During this introduction chapter, we have discussed the basics of semiconductor optical amplifiers and their characteristics and mode-locking for generating ultra-short pulses using semiconductor gain media. This dissertation is organized into five other chapters and four appendixes.

In Chapter II, novel amplification approach for high power semiconductor mode-locked

laser is discussed. In this chapter, new semiconductor master oscillator and power amplifier

17 approach, which is called eXtreme Chirped Pulse Amplification (XCPA) is discussed and

demonstrated. We build all semiconductor XCPA system from a master oscillator to an amplifier

system. As a master oscillator, a colliding pulse semiconductor mode-locked laser which is

author’s contribution in XCPA demonstration is presented and characterized. Using this novel amplification design, we achieved a record peak power in an all semiconductor mode-locked laser.

In Chapter III, a novel semiconductor mode-locked oscillator which is inherent from

XCPA system is presented. The characteristics of this novel semiconductor mode-locked

oscillator which is called an eXtreme Chirped Pulse Oscillator (XCPO) are investigated. XCPA

gain in the cavity, scaling properties of mode-locked spectrum, and low repetition rate

semiconductor mode-locked laser due to XCPO design are discussed. The newly developed high

power semiconductor amplifier is used to increase the power of the XCPO. A novel feed forward

noise cancellation scheme is discussed to suppress the intensity noise of the stretched pulse. This

noise cancellation scheme also plays role in gain flattening of the stretched pulses.

In Chapter IV, a functional device which is called Grating Coupled Surface Emitting

Laser (GCSEL) is introduced. For the first time, passive and hybrid mode-locking of GCSEL is

achieved by using saturable absorption in the passive section of GCSEL. To verify the present

limitation of the GCSEL for passive and hybrid mode-locking, a dispersion matched cavity is

explored. In addition to this, a Grating Coupled Surface emitting Semiconductor Optical

Amplifier (GCSOA) is also presented. An energy extraction experiment for GCSOA using

stretched pulses generated from the colliding pulse semiconductor mode-locked laser, which

shows XCPA advantages is also demonstrated. In this experiment, a dispersion matched

amplification scheme is used.

18 In Chapter V, another amplification method, passive optical cavity amplification or an enhancement cavity is presented. In order to achieve the interferometric stability, polarization- sensitive Hänsch-Couillaud Method is employed. The colliding pulse semiconductor mode- locked laser is also used as a master oscillator of picosecond optical cavity. The astigmatism free

optical cavity employed with an Acousto-optic modulator (AOM) is designed to a successive

accumulation of the picosecond optical pulses in the enhancement cavity.

Chapter VI summarizes this Dissertation, interesting areas of applications of which we have developed in this Dissertation, and a future direction for research.

In Appendix A, a simple source code for generating a transform-limited pulse from a

mode-locked spectrum is presented.

In Appendix B, a source code in ASAP for the astigmatism free enhancement cavity and a

Matlab source code for a transform-limited pulse from compressed pulses’ spectrum are

presented.

In Appendix C, a 2-Section semiconductor mode-locked laser is discussed as a master

oscillator.

In Appendix D, a novel frequency swept semiconductor laser using CFBG is presented.

19 CHAPTER 2

EXTREME CHIRPED PULSE AMPLIFICATION

2.1 High power semiconductor mode-locked laser

High power mode-locked semiconductor lasers are extremely attractive to commercial and defense use for a variety of applications, including signal processing, material processing, free space communication, range finding and medical tools. However, the gain saturation and the damage threshold of the facets of semiconductor optical amplifiers (SOAs) are the limiting factors for achieving high power mode-locked semiconductor lasers. In order to increase the output power from semiconductor mode-locked lasers, several approaches have been proposed and demonstrated, such as, surface emitting area structures including grating-coupled surface- emitting lasers (GCSELs) [1] and vertical-external-cavity surface-emitting lasers (VECSELs) [2], master oscillator and power amplifier (MOPA) system with tapered amplifiers [3] or inverse bow–tie amplifiers [4]. However, in these approaches high power operation approach is achieved by taking advantage of the increase in the active area. As a generic high power laser system, the

MOPA approach is very attractive due to the compactness and simple electrical pumping. In the case of short pulse amplification, however, the nonlinear carrier dynamics is strongly involved as stated in introduction.

High power short pulse semiconductor laser is simply achieved by using a mater oscillator and power amplifier system depicted in Fig. 2-1.

High Semi-MLL Power SOA

Fig. 2 - 1. Schematic of mater oscillator and power amplifier (MOPA) system. Semi-MLL:

semiconductor mode-locked laser.

2.2 Concept of extreme chirped pulse amplification

The eXtreme Chirped Pulse Amplification (XCPA) [5] is a novel master oscillator and power amplification (MOPA) system [3]. Unlike other MOPA systems such as the chirped pulse amplification (CPA) system [6] for solid state lasers or erbium-doped fiber lasers, the XCPA [5] is the amplification of a stretched pulse, whose time duration is longer than that of the upper state carrier life time of semiconductor gain media.

Typically carrier life time of semiconductor gain media with quantum well structures is less than a nanosecond. If the stretched pulse is longer than a nanosecond, then this amplification belongs in the XCPA regime, which is the distinct feature contrasting from the CPA. The carrier life time of EDFA or solid state gain media is typically millisecond region [7]. The carrier life time of the gain medium is very important parameter in laser applications.

21 In XCPA, we can get scalable amplification factor compared to the ratio between the stretched pulse duration and the upper state carrier life time. Because the upper state carrier life time is proportional to the gain of the amplification, if we stretch the pulse longer than the carrier life time, we can get its proportional gain factor depending on the ratio between the carrier life time and the stretched pulse duration.

Another salient feature of the XCPA system is that an amplifier in the XCPA system can be regard as a CW laser because the stretched pulses can be expanded to pulse interval. Hence any nonlinearity induced by the short pulse inside the semiconductor gain media can be eliminated.

Fig. 2 - 2. The eXtreme Chirped Pulse Amplification (XCPA). Semi-MLL: semiconductor mode-

locked laser, tp: stretched pulse duration.

The all semiconductor based XCPA consists of a master oscillator, optical amplifiers and a highly dispersive element which has positive and negative dispersion as shown on Fig.2-2. The ultra-short pulses from the master oscillator are stretched to time duration longer than the carrier life time of a semiconductor by either negative or positive dispersion component, and then the

22 pulses are amplified, similar to a CW laser. Finally, these stretched pulses are compressed by passing them through a component with dispersion opposite to the one of the previous case.

2.3 Colliding pulse semiconductor mode-locked laser

In order to demonstrate the XCPA system, a short pulse laser is crucial. Semiconductor mode-locked lasers take advantage of passive mode-locking for generating short pulses. A saturable absorber from the Naval Research Labs is employed in the master oscillator for generating femtosecond pulses [5]. The saturable absorber consists of about 75 periods InGaAs wells surrounded by AlGaAs barrier made from molecular beam epitaxy shown on Table 2-1.

The substrate is an n-Type GaAs and the cap is a p-type contact layer as shown in Fig. 2-3 (a).

The saturable absorber is designed to operate at 980nm with a transmission type. AR coatings are deposited on the both sides of the saturable absorber. The exitonic absorption is depicted at around 980nm in Fig. 2-3 (b).

Table 2 - 1. QW saturable absorber wafer structure.

Material Layer Thickness Doping Concentration

SiN AR coating 125nm

GaAs P-Contact 1µm p-doped 3x1018

GaAs Buffer 0.25um un-doped

Al0.35 Ga0.65 As Barrier 10nm x 75 layers

In0.18 Ga0.82 As QW 8.5nm x 75 layers

GaAs Buffer 50nm un-doped

GaAs n-contact 1µm n-doped 3x1018

SiN AR coating 125nm

(a)

24 1.0

0.8

0.6

0.4

0.2 Spectral Power(A.U)

0.0 900 920 940 960 980 1000 Wavelength(nm)

(b)

Fig. 2 - 3. Saturable Absorber(a) The schematics of Saturable Absorber [8], (b) exitonic

absorption at 970nm.

Due to the transmission type of the saturable absorber, we can make a colliding pulse

semiconductor mode-locked laser (CPSML) as a master oscillator. The shortest pulses have been

obtained by the technique of colliding pulse mode-locking in dye lasers [9]. The colliding effect

on the saturable absorber makes local interference effect so that the saturable absorption is easily

reached compared to non-colliding configuration. Sometimes the passive mode-locking using a

reflective saturable absorber is called self-colliding pulse mode-locking [10].

To make this possible, we devised 2 schemes of the CPSML as shown in Fig 2-4 (c) and

(d), where one was a linear cavity and the other a ring cavity version. We demonstrated the linear cavity as a master oscillator in the XCPA system.

25 Fig. 2-4 (a) [11] and (b) [12] depict a typical CPSML developed in the 1980s and (c) and

(d) describe a novel CPSML invented for XCPA demonstration. The bi-directional ring cavity

case, the difference between a gain medium and a saturable absorber is one quarter of the cavity

length so that short pulses enable to avoid colliding in the gain medium and are amplified in every half of cavity round trip time. The mode-locked frequency of this configuration is more than the 2nd harmonic of the cavity fundamental. Hence it is naturally a harmonic mode-locked

laser. The anti-resonant linear cavity shown in Fig. 2-4 (b) was invented to employ a linear

cavity. One of advantage of this configuration is that this cavity can be operated in the cavity

fundamental so that there is no super-mode effect even though the spatial hole burning from the

linear configuration occurs.

As shown in fig. (c), the linear version of CPSML has a saturable absorber in the middle

of cavity so that the pulses meet in the middle of cavity. In this setup, the CPSML operates at the

2nd harmonic of the repetition rate and needs the proper positioning of the gain medium for

hybrid mode-locking because the synthesizer input should match the pulses.

(a)

(b)

(c)

(d)

Fig. 2 - 4. Colliding pulse semiconductor mode-locked lasers. (a) Bi-directional ring cavity, (b)

anti-resonant linear cavity, (c) linear cavity, and (d) unidirectional ring cavity.

Another novel configuration for the CPSML is the unidirectional ring cavity shown in

Fig. 2-4 (d). In this scheme, the key feature is that optical length between the saturable absorber

and the last mirror should match the pulse’s interval. For simplicity, we demonstrate the linear

CPSML depicted in Fig. 2-5.

Fig. 2 - 5. Picture of colliding pulse semiconductor mode-locked laser.

As mentioned above, the position of a semiconductor gain media is important because the external RF signal phase should match the pulse repetition rate in hybrid mode-locking, otherwise the external signal plays role in attenuate or even disturb passive mode-locking. In this setup, the semiconductor gain media is place at the one sixth of the cavity length to support passive mode-locking.

The linear CPSML is biased by 300mA DC current and driven by an RF synthesizer to achieve hybrid mode-locking. In order to avoid the mechanical interference and to match the RF signal to the generated pulses, a 1mm inverse bow-tie SOA is placed at one sixth of the cavity length. The fundamental cavity frequency is 143MHz, and due to the colliding of the propagating

29 and the counter-propagating pulses, the repetition rate of the mode-locking is 285MHz. Fig. 2-6,

(b) depicts the CPML optical spectrum centered at 975.5nm with a 5nm bandwidth and (c) shows that the externally compressed pulse using free space dispersion compensator, whose FWHM is

0.88ps with the assumption of hyperbolic secant pulses.

(a)

-30 -35

-40 -45

-50

-55 -60

-65

Power(dBm) Spectral -70 -75 970 972 974 976 978 980

Wavelength(nm)

(b)

1.0 Pulse Width: 0.875 psec

0.5 FWHM 1.35 psec Power(A.U)

0.0 02468 Delay(ps)

(c)

Fig. 2 - 6. The CPSML’s performance. (a) Sampling scope trace, (b) mode-locked spectrum, and

The average power is more than 6mW directly from the oscillator and the pulses are pre-

chirped. The uncompressed pulse width is around 60ps, which is good for attenuating non-linear effect in fiber when XCPA is working.

31 2.4 Chirped fiber bragg grating for XCPA

A highly dispersive component is also crucial for an XCPA system. Usually grating pair or prism is used to control the dispersion in the cavity or outside of the cavity. However, the dispersion of a free space compensator is limited by the geometry [13].

In order to make a large dispersion compensator with compact size, the recently developed long length Chirped Fiber Bragg Grating (CFBG) is a best candidate. Different wavelengths are reflected on the different positions of CFBG that satisfy the Bragg condition so that it generates delay from the mode-locked pulse, eventually the pulse is stretched with the degree of group delay shown in Fig. 2-7 (a). Hence a long length CFBG, made by 3M, plays an important role in the XCPA system realization [14]. This long length CFBG has ~ 2000ps/nm group delay with 6nm bandwidth centered at 974.5nm, as shown in Fig. 2-7 (b). The reflective band-width matches the optical bandwidth of the CPSML. The short pulse from CPSML, which is 0.88ps, can be stretched to ~ 10ns by using the CFBG.

One of the advantages of using a CFBG as a dispersion compensator enable to be accessible via two ports and one can use simultaneously use a single CFBG as a pulse stretcher and compressor. This results in reduction of the group delay ripple effect which may be detrimental when compressing the stretched pulses. It also reduces the number of component, as well.

(a)

8000 Red First Port 6000 Blue First Port 4000

2000

-2000

-4000 Group Delay(ps) Group -6000

-8000 972 974 976 978 Wavelength(nm)

(b)

Fig. 2 - 7. Chirped fiber bragg grating (a) schematic, (b) group delay

33 2.5 Extreme chirped pulse amplification system

Kim et al. demonstrated the XCPA system by using the CPSML as a master oscillator, a long CFBG having 2000ps/nm group delay as a stretcher and compressor and a tapered semiconductor optical amplifier in main components [5].

The main experimental setup is depicted in Fig. 2-8. The CPSML laser which shows more than 60ps, which is up-chirped, is stretched in a 980nm single mode fiber (HI1060) to reduce the nonlinearities of the CFBG when the pulses are compressed. The length of HI1060 is around 300meters. In order to avoid stretched pulse overlapping, a pulse picker employing the semiconductor gain medium is used. This pulse picker is utilized absorption in the reverse biased.

Hence the repetition rate of the mode-locked pulse train is 95MHz, which is 10.5ns time interval.

A polarization sensitive circulator consisting of a polarization beam splitter (PBS) and a Faraday rotator is used to change the direction of the reflected light from the CFBG. The stretched pulse from the CFBG is inserted in a ridge wave guide single mode SOA. The ridge-wave guide single mode SOA is used as a pre-amplifier to saturate an amplifier whose structure is taped. An optical band pass filter having 10nm bandwidths used to suppress the transmitted light from the blue port of the CFBG. After the final amplification of the tapered amplifier, the stretched pulse is compressed in the CFBG. Another polarization sensitive circulator is used to divert the light path to a free space dispersion compensator. In order to match the polarization of the components a half wave plates (HWP) are used to change the linear polarization.

RWGSOA Pulse Picker CPMLL OI prestretcher Beam shaper

BPF HWP PBS TA RWGSOA

PC PC PBS CFBG Beam FR shaper lens

HWP grating compressor

Fig. 2 - 8. eXtreme Chirped Pulse Amplification Setup [5].

CPMLL: Colliding pulse mode-locked laser, OI: Optical Isolator, RWGSOA: Ridge-wave guide semiconductor optical amplifier, prestretcher: HI1060 single mode fiber spool, PBS: Polarization beam splitter, FR: Faraday rotator, HWP: half wave plate, TA: Tapered semiconductor optical amplifier, PC: Polarization controller, CFBG: Chirped fiber bragg grating.

35 This XCPA system stretch less than 1ps pulse to 9.6ns which is substantially longer than the carrier life time of the SOAs which is around 1ns. The stretched pulse almost fills the repetition interval. Hence the pre-amplifier and the tapered amplifier increase power like CW amplification.

The XCPA system successfully compressed 9.6ns stretched pulse to 0.6ps as shown in

Fig. 2-9 and generates 1.4KW peak power which is a record for an all semiconductor mode- locked system [5]. Even in stretched pulse amplification, the XCPA system gives 1µJ energy extraction with pulsed bias condition.

No pulse overlapping

time

Wavelength

(a)

(b)

Fig. 2 - 9. The XCPA performance [5]. (a) Spectrally resolved streak camera trace of the

stretched pulse, (b) compressed autocorrelation.

2.6 Summary

In conclusion, we devised 2 kinds of the colliding pulse mode-locked semiconductor

lasers, where one is a linear cavity version and the other is a ring cavity version as master oscillators in the eXtreme Chirped Pulse Amplification system. The demonstration of the linear colliding pulse semiconductor mode-locked laser used in a master oscillator for realization of the

XCPA system and its generation of pulses with 0.88ps pulse duration and 285MHz repetition rate was also done.

37 The XCPA system generates 1.4kW peak powers, which is record peak power in all semiconductor mode-locked lasers and demonstrates 9.6ns stretched pulses and 0.6ps compressed pulses at 95MHz. In the stretched pulse amplification using pulse biased, more than

1µJ is achieved.

This new MOPA system in short pulse amplification reduces nonlinear effects in semiconductor gain media using quasi-CW amplification. Even it opens the possibility to use the stretched pulse for signal processing.

38 CHAPTER 3

EXTREME CHIRPED PULSE OSCILLATOR BASED ON THETA

CAVITY DESIGN

3.1 A more of extreme chirped pulse amplification

This chapter describes a novel extreme chirped pulse laser oscillator based on the XCPA

system that generates a periodic linearly chirped CW output and ultra-short pulses for generating

high pulse energy and signal processing applications. This oscillator scheme overcomes the

limitations of semiconductor mode-locked lasers by stretching and compressing pulses inside the

cavity. As stated in the introduction, due to the carrier dynamics of the semiconductor gain media, it is very difficult to achieve an ultra-short pulse out of the oscillator. Hence temporal breathing in the cavity is useful concept in order to achieve ultra-short pulse or high power in semiconductor gain media.

A ‘free-space’ breathing mode dispersion managed mode-locked diode ring laser that uses the idea of intra-cavity temporal pulse shape management to minimize nonlinearities in the semiconductor gain media has been demonstrated [1, 2]. The extreme chirped pulse mode-locked oscillator is an evolution of the dispersion managed oscillator, but with a new key advantage.

An oscillator version of XCPA, which is called XCPO, reveals the advantages of the

XCPA approach to achieve high power and mitigate the non-linear effects more clearly. In this operation regime, the mode-locked pulses are temporally stretched to a temporal duration that equals the pulse repetition period, converting the pulse train into a CW signal by using a long-

39 length chirped fiber bragg grating [2]. By amplifying this CW signal in the semiconductor

optical amplifier, the detrimental effects of transient gain saturation and integrating self phase

modulation are eliminated [4, 5].

In order to investigate how the stretched pulse whose time duration is longer than the

carrier life time of semiconductor gain media, the saturation power in the amplifier should be

considered. A Master oscillator and power amplifier system is used to increase output power of

semiconductor gain media however. It has a gain saturation which limits the maximum power

that can be extracted from the amplifier. Due to this gain saturation, the saturation power is

defined. The saturation power can be considered as the maximum output power of mode-locked

semiconductor Master oscillator and power amplifier systems. The saturation power may be

expressed as [6],

hA ν [3-1] Ps = Γ α τ ,

where, Ps is the saturation power, hν is the photon energy, A is the active region area, Γ is the optical confinement factor, α is the differential gain and τ is the gain recovery time. In order to increase the saturation power, one can increase the area of the active region or use quantum dot gain media whose gain recovery time is shorter than quantum well gain media.

Unlike other approaches which take advantage of the active area, our approach uses a swept frequency or a stretched pulse to extract more pulse energy. If the stretched pulse duration is longer than the upper state carrier life time, one can extract more energy from the SOA and obtain a scalable energy extraction factor based on the power saturation with respect to stretched pulse duration as shown in following equation,

40 t [3-2] E ∝ p τ ,

where, E is the energy extraction, tp is the stretched pulse duration in the SOA and τ is the gain recovery time of the SOA. We refer to this as the XCPA gain effect.

This approach is also successful in terms of mitigating nonlinearities of the SOA as well.

In this regime, the mode-locked pulses are temporally stretched to a temporal duration that equals the pulse repetition period, converting the pulse train into a CW signal by using a long- length chirped fiber bragg grating (CFBG). By amplifying this CW signal in the SOA, the detrimental effects of transient gain saturation, such as carrier heating and cooling, and integrating self phase modulation are eliminated. Since the gain dynamics of the semiconductor optical amplifier operate in the CW injection regime, the limitation to the extracted pulse energy under transient optical amplification can be circumvented [5].

3.2 Preliminary experiment for extreme chirped pulse oscillator

In order to implement XCPO, first, we use the 2 CFBG having slightly different group delay. It is very difficult to fabricate the same group delay CFBGs, especially large group delay

CFBGs. The group delay of one CFBG is 542ps/nm and the other is -535ps/nm. The bandwidth of two CFBG is around 7~8nm centered at 1563nm. Two CFBGs are inserted in the semiconductor ring laser. Mode-locking is accomplished with loss modulation of a 10GHz intensity modulator. Two output couplers are placed to monitor the stretched pulse and the short pulse out of cavity. The schematic of the mode-locked laser is shown in Fig. 3-1.

41 SC1 90:10 SOA

PC OSA CFBG CFBG Cir Cir SC2 542ps/nm -535ps/nm PC Mod 90:10 G

Sig

Fig. 3 - 1. Preliminary setup for eXtreme Chirped Pulse Oscillator.

SOA: Semiconductor optical amplifier, CFBG: Chirped fiber bragg grating, Mod: Intensity modulator, Sig: Signal generator, PC: Polarization controller, SC1: Sampling scope 1, SC2:

Sampling scope 2, OC: Output coupler, OSA: Optical spectrum analyzer, AC: Auto-correlator.

There is also a small transmittance from the CFBGs that is also monitored. In order to

measured short pulses, an intensity auto-correlator is used after amplification by an erbium-

doped fiber amplifier (EDFA). The stretched pulse is monitored using a sampling scope and an

optical spectrum analyzer. Due to the circulators in the cavity, an optical isolator is not needed.

42 0.006 0.018 0.016 0.014 0.004 0.012 0.010 0.008 0.006 0.002 Power(A.U) Power(A.U) 0.004 0.002 0.000 0.000 50200 50400 50600 50800 51000 50200 50400 50600 50800 51000

Time(psec) Time(psec) (a) (b)

Fig. 3 - 2. Performance of oscillator. (a) Sampling scope trace of compressed pulse, and (b)

sampling scope trace of stretched pulse.

The mode-locked frequency is 2.2 GHz, which is harmonic mode-locking. The cavity

fundamental is around 44MHz. The output power of the compressed port is 1.64mA at 600mA

biased current to SOA. As shown in Fig. 3.2 (a), the compressed pulse shows a clean sampling

scope trace. The stretched pulse is also monitored on sampling scope. As shown in stretched

pulses on sampling scope, the stretched pulse is operated in CW mode. The stretched pulse in

CW is also repeated every 450ps. The energy fluctuation with respect to the spectra in a

stretched pulse plays a role in intensity noise of CW laser.

In this experiment, the 2.2GHz repetition rate which is ~ 500ps time widow for loss

modulation is short enough in terms of time window for mode-locking. However, if the

repetition rate is reduced, short time window for mode-locking is needed such as an electrical

comb generator or passive mode-locking mechanism.

1.0 1.0 RBW:0.05nm

0.8 0.8

0.6 0.6 12.8psec 0.4 0.4

SHG A.C (A.U) 0.2 0.2

Spectral Power(A.U)

0.0 0.0 1560 1562 1564 1566 Delay(32ps/Div) Wavelength(nm)

Fig. 3 - 3. Performance of oscillator. (a) Intensity auto-correlation, and (b) mode-locked

bandwidth.

In order to further investigate the short pulses, second harmonic intensity auto-correlator

is used. The width of autocorrelation is 12.8ps, which is 8.3ps assuming a hyperbolic secant

pulse. In hyperbolic secant pulse, de-convolution factor is 1.54. When considering the stretched

pulse width, the breathing ratio in the cavity is 54 which are defined by the stretched pulse width

divided by the short pulse width.

The mode-locked spectrum bandwidth is ~1nm centered at 1562.7nm which is close to

the center wavelength of CFBGs. When considering the group delay of the stretched CFBG, we

can get 540ps stretched pulse. However, the repetition time interval is ~450ps so that the

stretched pulse is a little bit overlapped in time domain. The difference between the overlapped

spectra is large enough not to be monitored using the interference in electrical response of a fast

photodetector or in the resolution bandwidth of optical spectrum analyzer.

44 3.3 Theta cavity design for extreme chirped pulse oscillator

From preliminary experiment for XCPO, we have learned two problems.

z Matching group delays of two different CFBGs.

z Short time window for mode-locking.

Since it is very difficult to match the group delays of two different CFBGs, using a single

CFBG is the best solution for matching the group delay. In this case, the group delay ripple can be significantly reduced because we use the same CFBG for stretching and compressing.

However, residual transmission of outside bandwidth or even inside bandwidth of CFBG will cause problems.

The theta laser schematic is shown in Fig. 3 - 4. An SOA is inserted in an external cavity incorporating a CFBG, having a dispersion of ±340 ps/nm, and a 20 GHz modulator. A single

CFBG is used as a stretcher and a compressor in order to minimize the total group delay ripple.

This design is a solution for problems which have occurred in cavity having two CFBGs.

The measured reflection and group delay for each port are shown in Fig. 3-5. It shows a >

10 dB difference between the reflection and transmission, and the opposite group delay from each port of a single CFBG. We use a low loss 20 GHz modulator for active mode-locking and a tunable band pass filter having a 5 nm FWHM to prevent closed loop CW lasing due to a leakage of transmitted power from the CFBG.

BPF PC SOA OC

CFBG

Mod PC OC

Electric Comb Generator

Fig. 3 - 4. Experimental configuration for the extreme chirped pulse mode-locked diode laser

ring oscillator.

SOA: Semiconductor Optical Amplifier; CFBG: Chirped Fiber Bragg Grating; C1, C2: Optical

Circulator; Mod: Intensity Modulator; OC: Output Coupler; BPF: Band Pass Filter; PC:

Polarization Controller; T: Pulse Repetition Period.

2000 -10 1500 1000 -15 500

0 -20

-500 -25 -1000 Delay(ps) Group

-1500 Reflectance -30 Transmittance Relative R and T(dB) -2000 1548 1550 1552 1554 1556

Wavelength(nm)

2000 1500 -10 1000 -15 500

0 -20

-500 -25 -1000

Delay(ps) Group

-1500 Reflectance -30 Relative R and T(dB) Transmittance -2000 1548 1550 1552 1554 1556

Wavelength(nm)

Fig. 3 - 5. Relative reflectances, transmittances and group delays of the two ports of the CFBG.

0.1

0.0

-0.1

20psec -0.2

Voltage(V)

-0.3

500MHz/10dBM -0.4 51.5 51.6 51.7 51.8 51.9 52.0 52.1

Time(ns)

Fig. 3 - 6. Temporal response of an electric comb generator.

An electrical comb source is utilized to provide a short time window for active mode- locking. This electrical comb source has 50GHz electrical bandwidth so that the temporal pulse width is around ~20ps as shown in Fig. 3-6. The electrical pulse is amplified by use of a broadband modulator driver to achieve a sufficient modulation index. Because the directivity of the circulators is more than 50 dB, an optical isolator for unidirectional operation is not needed.

The laser is modulated at 1.95 GHz, generating periodic pulses separated by 510 ps.

Optical pulses are formed by the modulator after passing the output coupler, and are directed to the CFBG. The reflected pulses are temporally stretched to 510ps and are passed through the optical filter and polarization controller. The CW stretched pulse train is then amplified in the

SOA, where a second output coupler samples the stretched pulse train. The stretched pulses are

48 then directed to the CFBG, where they are reflected and compressed to 3.9ps and then reshaped in the modulator.

Power(A.U)

0.0 0.5 1.0 1.5 2.0

Time(nsec)

Fig. 3 - 7. Sampling scope image of compressed pulse at 1.95GHz.

The fundamental cavity frequency is 6 MHz and the synthesizer is tuned to the 325th harmonic of the fundamental frequency to achieve harmonic mode-locking. To demonstrate the operation of the mode-locked oscillator, Fig. 3-7 shows pulse trains monitored with a 20 GHz sampling scope. Well defined pulses are observed with a pulse period of 510ps. When driving the intensity modulator with the electrical comb generator, the modulator’s bias point is properly chosen to achieve short pulses. Because the electrical comb generator voltage is negative, the intensity modulator’s Q point is chosen on the downward slope of the transmission with respect to applied voltages.

1.0 600mA 400mA

0.8 200mA Transform -limited 0.6

0.4

0.2

Autocorrelation Signal(A.U) Autocorrelation 0.0 -30 -20 -10 0 10 20 30 40 Delay(ps)

Fig. 3 - 8. Autocorrelation of 200mA, 400mA, 600mA and transform-limited pulse at 1.95GHz.

The autocorrelation evolution with respect to increasing current is shown in figure Fig. 3-

8. Because of the CW nature of the stretched pulse in the gain medium, the current can be increased without concern of detrimental nonlinearities associated with the gain medium. The autocorrelation FWHM of the compressed pulse from the oscillator at biased 600mA is 5.6ps.

The pulse width of 3.9ps at 600mA is calculated from a de-convolution using the spectral intensity to calculate the intensity autocorrelation. The experimentally measured autocorrelation is 1.5 times the transform limit. In order to better understand the pulse evolution, autocorrelation traces for different currents are traced at different time delays.

1.0

0.8

0.6

5.5psec 0.4

SHG A.C (A.U) 0.2

0.0 Delay(16psec/div)

Fig. 3 - 9. Externally compressed auto-correlation of the compressed pulse.

The compressed pulse from the oscillator is compressed to shorten pulse width outside of the oscillator using telescopic grating compressor which is not shown in experimental setup. The

auto-correlation width of the external compressed pulse is 5.5ps, which is very close to the

uncompressed auto-correlation. We already calculated the time bandwidth product from the

uncompressed pulse. Hence it is obvious that all chirping is compensated in the CFBG without

higher order chirping so that a transform-limited auto-correlation from the mode-locked

spectrum can be calculated, assuming no dispersion. The details of how we calculate the transform-limited pulse from the mode-locked spectrum will be discussed in the Appendix.

1.0 80

60 40

20 0 0.5

-20 Power(A.U) -40 -60 Chirp Frequncy(GHz) Chirp -80 0.0 0.0 0.5 1.0 1.5 2.0

Time(nsec)

Fig. 3 - 10. Time-resolved chirp measurement and sampling scope image of stretched pulse.

Fig. 3-10 shows the time-resolved chirp (TRC) and sampling scope trace of the stretched pulse. The TRC is measured with an Agilent TRC measurement system consisting of an Agilent

86146B optical spectrum analyzer and an Agilent 86100B sampling scope. The chirp measurement shows a 154.9 GHz chirp centered at 1551 nm.

The inverse slope of the TRC measurement data of 330ps/nm agrees well with the

dispersion of the CFBG, 340 ps/nm. TRC measurement system will be discussed in Appendix.

1.0

0.8

0.6

0.4

0.2 Spectral Power(A.U) (c) 0.0 1548 1550 1552 1554

Wavelength(nm)

Fig. 3 - 11. Mode-locked spectrum of stretched pulse.

The 1.5 nm optical bandwidth of mode-locked spectrum for the stretched pulses is shown in Fig. 3-11. Considering the dispersion of the CFBG, this spectral width implies pulses of 510 ps. However, there is power in the wings of the spectrum extending over 2 nm, corresponding to

680 ps pulses. This implies an overlap of the red end of the spectrum from one pulse with the blue end of the spectrum from the neighboring pulse. Hence, the TRC measurement cannot resolve the overlapped spectrum as a chirp frequency. Despite this spectral overlap, note the linearity of the group delay and the sampling scope image of the stretched pulses with less than

50% modulation. The temporal profile of the stretched pulses also looks similar, as shown in the sampling scope trace and the optical spectrum in Fig. 3-10 and 3-11, which demonstrates the wavelength to time mapping in highly large dispersive media [7].

Compressed Pulse 12 Stretched Pulse 10

Power(mW) 4

0 100 200 300 400 500 600 Current(mA)

Fig. 3 - 12. L-I curve for stretched pulse and compressed pulse output ports with 10%

transmission output couplers, respectively.

Because the stretched pulse acts as a CW light source in the SOA, the L-I characteristics are measured as shown in Fig. 3-12. At 600mA, we achieve 13.3mW average power from the stretched pulse port and 1.3 mW from the compressed pulse port, with 10% transmission output couplers. The main reduction of power from the compressed port is due to the loss of the modulator and the CFBG. The threshold current is ~150mA and the pulses maintain their general characteristics spanning a pumping current range of ~4X the threshold current. In this setup, the loss of the cavity, ~10dB, and the low saturation power of the SOA are the main limitations for high power operation. We believe that if the loss of the CFBG is improved and a high power amplifier is used, several hundred mWs in average power could be achieved.

54 In conclusion, we have demonstrated a novel optical cavity design to realize an extreme

chirped pulse mode-locked laser. This novel design, or theta cavity, generates near transform

limited 3.9ps mode-locked optical pulses and a periodic 510ps linearly chirped CW output with

1.5nm mode-locking bandwidth. The novel cavity design is used to overcome the limited energy

extraction and nonlinearities induced by the transient gain dynamics in mode-locked

semiconductor lasers. In this experiment, we obtain 13.3mW of average power at 600mA from

the stretched port. These results suggest that the extreme chirped pulse mode-locked oscillator

based on semiconductor gain medium and theta cavity design may be useful for ultra-short, high

output energy pulse generation.

3.4 Extreme chirped pulse oscillator for > 1ns stretched pulses

We proposed and demonstrated eXtreme Chirped Pulse Amplification (XCPA) [8] and an

eXtreme Chirped Pulse Oscillator (XCPO) [9] to reduce detrimental nonlinearities caused by

SPM, carrier cooling and heating, and to increase the output power from semiconductor mode- locked lasers. These approaches are different from the high power pulse generation approaches increasing an active area of an optical amplifier. The XCPA and the XCPO are temporal power scaling approaches unlike the spatial scaling approaches. In our previous results, we obtained record peak powers from the XCPA and 520ps CW-like linearly frequency-swept output pulses from the mode-locked XCPO.

In this section, we demonstrate 10ns frequency-swept pulses which are substantially longer than the upper state carrier life time of the SOA, and a 100MHz repetition rate which is

55 the lowest repetition rate semiconductor mode-locked laser with loss modulation while

maintaining the beneficial characteristics of the XCPO. This direct stretched pulse generation

from the oscillator can be used in applications such as, swept-source optical coherence

tomography (OCT) [10], and time stretched photonic analog to digital converters (P-ADC) [11].

3.4.1 Theta cavity and experimental setup

In this work, we generate a 10ns stretched pulse directly from the oscillator, noting that

this temporal duration is substantially longer than the carrier life time (~1ns) and represents a 19

fold improvement compared to our previous work [9]. In this operating regime, the stretched

pulse duration enables one to observe the XCPA gain in the oscillator.

The use of a Theta cavity design operating in the XCPA regime in semiconductor gain media opens the possibility that with increased temporal stretching, where the pulse duration is longer than the carrier life time, one can extract more pulse energy in which the pulse energy is linearly proportional to the ratio of the stretched pulse duration to the carrier life time. In addition, one can obtain a scalable mode-locked spectral width with respect to pulse repetition rate.

The tendency that the temporal duration of the stretched pulse equals the pulse repetition period forces the spectral width to scale in terms of the repetition period due to the large group delay of the CFBG. This characteristic is understood as a spectral filtering of frequency-swept pulse train to avoid overlapping of the stretched pulses.

56 The performance of CFBG is summarized in Fig. 3-13. The bandwidth of the CFBG is C- band and insertion loss is less than 3.5dB. It should be noted that the group delay data out of

1570nm is artifact of the measurement system in Fig. 3-13 (a). In Fig. 3-13 (c), there is a modulation periodicity in group delay ripple. This is due to the method used for inscribing lines during the CFBG fabrication. CFBGs cannot be together in one continuous step. The grating is written in sections of at most 35cm. After each section the fiber is feeded forward and the grating is continued were the last section ended. The periodicity of these sections is evident in our measurement traces. For fibers with higher dispersion these stitched are located more closely in wavelength.

Insertion Loss for Two Ports 10

0 Insertion Loss(dB) Insertion -2

1520 1530 1540 1550 1560 1570 1580

Wavelength(nm) (a) (b) GDR1+-GDR2 Group Delay Ripple for Two Ports

60 10

40 5 20 0 0

GDR(ps)

GDR(ps) -20 -5 -40

-60 -10

1520 1530 1540 1550 1560 1570 1520 1530 1540 1550 1560 1570 Wavelength(nm) Wavelength(nm)

Fig. 3 - 13. Performance of the 2000ps/nm CFBG. (a) Group delay, (b) insertion loss, (c) group

delay ripple, and (d) sum of group delay ripple.

Even though one CFBG is used to match the group delay and to reduce the group delay ripple in the Theta cavity, there is still some residual group delay ripple in the CFBG as shown in

3-14 (d). This is the sum of group delay ripple from the red port and the blue port of the CFBG.

58 When considering the residual group delay of stretched pulse and compressed pulse, this

uncompensated group delay will be a limitation of the Theta cavity. In laser oscillator, multiple round trips accumulate the uncompensated group delay.

An electrical comb generator’s performance is measured in a sampling scope as shown in

Fig. 3-14. The frequency of a 100MHz sine wave with -2dBm input generates an electrical comb in the time domain with a 100MHz step recovery diode. Vertical axis is voltage on 100mA/div with 30dB attenuation. Hence no additional RF amplifier is needed.

Fig. 3 - 14. Electrical comb generator at 100MHz.

The XCPO implemented with a Theta cavity schematic is depicted in Fig. 3-15. An SOA

is inserted in an external cavity incorporating a CFBG, having a group delay of 2000 ps/nm, and

a 10GHz modulator. The single CFBG is used as both the stretcher and the compressor in order

to minimize the total group delay ripple like previous experiment. Because the difference

between the transmittance and the reflection of the CFBG is only 8dB, a polarizer with a

polarization controller is inserted to block any residual transmitted light. An optical band pass

59 filter with a 10nm bandwidth is used to remove the transmitted light outside of the bandwidth of

CFBG.

The enclosed box for the Theta cavity is shown in Fig. 3-16. The box supports

environmental isolation.

Fig. 3 - 15. Diagram Theta cavity for Extreme Chirped Pulse Oscillator (XCPO) made with an

SOA.

CFBG: Chirped Fiber Bragg Grating, SOA: Semiconductor Optical Amplifier, Mod:

Intensity Modulator, OC: Output Coupler, PC: Polarization controller, POL: Polarizer,

BPF: Optical Band Pass Filter, T: Repetition Rate.

Fig. 3 - 16. Enclosed box for eXtreme Chirped Pulse Oscillator.

Active mode-locked operation via loss modulation is performed as follows: an intensity modulator incorporating an electric comb generator produces a short optical pulse, and then it is stretched in the CFBG. The SOA amplifies the CW-like mode-locked pulse using a DC-bias. The chirped and amplified light is compressed by the same CFBG, and reinserted into the intensity modulator. A single CFBG is used for reducing the group delay ripple (GDR) that would be caused from the mismatch of group delays if two independent CFBGs are used. Two output couplers with 10% output coupling ratio are inserted into the cavity to monitor the performance of the stretched and compressed pulses.

61 3.4.2 Experimental results

The performance of the stretched or frequency-swept pulse is summarized in Fig. 3-15.

The spectrally resolved streak camera traces and their mode-locked spectra are depicted in Fig.

3-17 for 3 repetition rates of the XCPO. It can be seen that the CW-like output fills the repetition

interval by broadening the mode-locked spectrum. The grating used in the spectrally resolved

streak camera measurement is 500g/mm. In this data, the down chirped pulse completely spans

the pulse period, which means that as the repetition rate is reduced, the mode-locked optical bandwidth becomes linearly proportional to the repetition interval for a given group delay of the

CFBG. As mentioned earlier, ultra-fast carrier dynamics and the integrated SPM effect in the

SOA are completely eliminated due to CW injection into the SOA. The stable 102MHz stretched pulse result shows that the theta cavity design overcomes the fast gain dynamics in the SOA.

62 0

-10

-20

-30 Spectral Power(dBm)

1546 1548 1550 1552 1554 1556 1558 Wavelength(nm)

(a) (b)

-10 -10 -20 -20 -30

-30 -40

-40 -50 Spectral Power(dBm) Spectral Power(dBm) Spectral

-50 1540 1545 1550 1555 1542 1544 1546 1548 1550 1552 Wavelength(nm) Wavelength(nm) (c) (d)

Fig. 3 - 17. The stretched pulse performance of the Theta cavity. (a) Spectrally resolved streak

camera image, (b) mode-locked spectrum at 505MHz, (b) 246MHz, and (c) 102MHz.

All data are taken at 600 mA DC-biased SOA.

30 4 25

2 15 10

y=A+B*X

Spectral Width(nm) 5 A=0.09022

B=0.45924 Energy per Pulse(pJ) 0 246810 0 (ns) 246810(ns) 1/(Repetition Rate) 1/(Repetition Rate)

(a) (b)

Fig. 3 - 18. Scalability of pulse energy and spectrum. (a) Mode-locked spectral bandwidth vs.

repetition period, and (b) Energy per pulse for 505MHz, 246MHz, and 102MHz

repetition rate.

The slope of Fig. 3-18 (a) matches the group delay of the CFBG in this experiment, which is 2000ps/nm, e.g., note the same slope of the stretched pulse in the spectrally resolved streak camera trace. The output pulse energy is summarized in Fig. 3-18 (b). At 102MHz repetition rate, a 33pJ energy is obtained directly from the oscillator with a 10% output coupler.

The linear slope of Fig. 3-18 (b) implies that the average power of the oscillator is maintained, owing to the CW injection into the SOA. As mentioned earlier, one can obtain a scalable energy extraction with respect to the stretched pulse width. In addition to the scalable energy extraction from the oscillator, the scalable mode-locked bandwidth due to the large group delay of the

CFBG and repetition rate of the Theta cavity is a distinct characteristic of the Theta cavity.

64 2ns

(a) (b)

Fig. 3 - 19. Sampling scope traces at (a) 246MHz, and (b) 102MHz.

The short pulse performance of the Theta cavity operating at 246MHz and 102MHz is summarized in Fig. 3-19 (a, b). The sampling scope trace shows a clean pulse train and the intensity autocorrelation as shown in Fig. 3-20 shows a 3.6ps pulse width, estimated by the 2.0 deconvolution factor which is calculated from the mode-locked spectrum. The autocorrelation width is measured to be 1.8 times larger than the transform-limited pulse, considering the calculated autocorrelation. It should be noted that low repetition rate operation, such as 100MHz, is generally very difficult to be realized, in semiconductor mode-locked lasers owing to the fast gain recovery of the semiconductor gain medium, especially for loss modulation, however, the

XCPO approach provides a solution for low repetition rate semiconductor mode-locked lasers and high power operation while maintaining other key advantages of semiconductor gain media.

Moreover, it should be noted that a further reduction in repetition rate can be achieved by increasing the group delay of a long CFBG.

In order to generate the shortest pulse at 102MHz, the bias current of SOA is reduced to

250mA. This reduces the mode-locked bandwidth to ~1nm. This is a primarily due to the CFBG

65 which limits the performance of the theta cavity. Even though a single CFBG is used to reduce the GDR of the CFBG, the fabrication process of the CFBG, such as single-side inscription produces birefringence effects in the CFBG [13] which leave a residual GDR in the Theta cavity even though a single CFBG is used. We believe that this limits the temporal duration of pulses generated from the Theta cavity.

1.0 1.0 Trasform-limited AC 0.8 Measured AC 0.8 0.6 0.4 0.2

Spectral Power(A.U) Spectral 0.0 0.6 1554 1557 1560 1563 Wavelength(nm) 7.1ps

0.4

SHG(A.U)

0.2

0.0 -40 -20 0 20 40 Delay(ps)

Fig. 3 - 20. Second harmonic intensity autocorrelation at 102MHz. Inset is the mode-locked

spectrum for generating short pulses.

66 In order to increase the mode-locked bandwidth and increase the output power, we use a

660ps/nm group delay CFBG, a 20nm optical band pass and a 20% output coupler after the SOA.

The CFBG has less than 3dB insertion loss over C-band and the group delay of the CFBG has completely opposite slope depending on the input port from the CFBG as shown in Fig. 3-21 (a).

The reflection from the blue port shows a higher reflection than that from the red port in the

shorter wavelength due to the cladding modes. The sampling scope trace of the compressed port

is shown in Fig. 3-21 (b) which depicts a clean pulse train at 99MHz.

The stretched pulse in the spectrally resolved streak camera data shows that the spectrum

fills the repetition interval in Fig. 3-22 (a). The grating used in spectrally resolved streak camera

trace is 50g/mm. In this case, the mode-locked bandwidth is 14.6nm (10dB) as shown in Fig. 3-

22 (b). Hence the stretched pulse duration is estimated to be ~9.6ns (10dB) due to the

wavelength and time mapping of the stretched pulses. The pulse energy from the stretched pulse

is 58.4pJ directly from the oscillator

20 0

15 -2 10

5 -4

0 -6 -5

Group Delay (ns) -10 -8 Refelectance (dB) Refelectance -15 -10 1530 1540 1550 1560

Wavelength (nm) (a) (b)

Fig. 3 - 21. 660ps/nm CFBG and short pulse. (a) Group delay and Reflection, (b) sampling

scope trace.

(a)

-20

-30

14.6nm -40 (10dB)

-50

Spectral Power(dBm) -60

1540 1550 1560 1570 1580

Wavelength(nm) (b)

Fig. 3 - 22. eXtreme Chirped Pulse Oscillator with 660ps/nm CFBG. (a) Spectrally resolved

streak camera of the stretched pulse and (b) mode-locked spectrum.

In conclusion, we have demonstrated a mode-locked eXtreme Chirped Pulse Oscillator

(XCPO) implemented with a Theta cavity design that generates 10ns linearly chirped pulses and

compressed pulses of 3.6ps in duration while maintaining various key advantages. In this

experiment, two distinct advantages of the Theta cavity, which are the high pulse energy of

semiconductor gain media and the mode-locked spectrum scalability, are demonstrated. By

using these characteristics, we obtained a pulse energy of 58.4pJ from the stretched pulse and a

mode-locked optical bandwidth of 14.6nm (10dB) directly from the oscillator at a repetition rate

of 99MHz. The oscillator can be used for high speed frequency-swept optical coherence

tomography (OCT) and time-stretched photonic analog digital converters (P-ADC). As a new

regime in semiconductor mode-locking, future directions will focus on expanding the mode-

locked bandwidth and to use high power semiconductor amplifiers in the gain medium.

3.5 Amplification of stretched pulses from the XCPO

In order to increase, the wall plug efficiency and increase the optical power, recently

developed high power SOAs called SCOWAs (Slab-Coupled Optical Waveguide Amplifiers)

[12,13]], have been used. One of advantage of SCOWAs is the high saturation power which is good for stretched pulse amplification. Due to this reason, SCOWAs are good for high power oscillators and ultra-low noise semiconductor mode-locked lasers. The gain data with respect to

current of a packaged SCOWA device that was used in the experiment is shown in Fig. 3-23

(a,b).

(a) (b)

Fig. 3 - 23. Slab-Coupled Optical Waveguide Amplifier. (a) Measured gain, and (b) SCOWA

package.

The SOWA has 25dB gain at 3A so that it is a good candidate as a gain medium for the

Theta cavity to get higher powers directly from the oscillator. The saturation power at 3A is

~22dBm.

2.0

1.5

1.0 Voltage(V) 0.5

0.0 0 1000 2000 3000 4000 Current(mA) (a) (b)

Fig. 3 - 24. SCOWA characteristics. (a) Amplified spontaneous emission, and (b) I-V curve.

250

200

150

50ns 100

Power(mW) 50 ~ 5nm 0 1.0 1.5 2.0 2.5 3.0 3.5 Current(A)

(a) (b)

Fig. 3 - 25. SCOWA amplification. (a) Spectrally resolved steak camera trace, and (b) increased

power vs. current.

The Theta cavity is operated at 106MHz of repetition rate using 2000ps/nm CFBG. The average power of the stretched pulses is 4.4mW. The stretched pulses are injected to SCOWA via a polarization controller.

The maximum output power at 3.5A is 246mW as shown Fig. 3-25. The slope efficiency of amplified system is 0.92W/A and the spectrally resolved streak camera trace is shown in Fig.

3-25 (a) after amplification. It should be noted that the amplification of the stretched pulse shows gain saturation effect of SOA when considering the mode-locked spectrum as shown in Fig. 3-26.

This effect can be used for AM noise cancellation of the stretched pulses or gain flattening because of the wavelength and time mapping of stretched pulses.

-20 -20

-30 -30

-40 -40 Spectral Power(dBm) Spectral Power(dBm) -50 -50 1545 1550 1555 1560 1545 1550 1555 1560 Wavelength(nm) Wavelength(nm) (a) (b)

Fig. 3 - 26. Amplified spectrum. (a) The stretched pulse spectrum of the Theta cavity, and (b)

after SCOWA amplification.

This Theta cavity and SCOWA configuration is a good candidate for generating high power pulses with a good wall-plug efficiency (WPE). The calculated WPE values are summarized in Table 3-1. In this calculation, the total driving powers for a synthesizer, a current driver, and a temperature controller are not included, however, the consumed powers for a SOA, a TE cooler and modulator are considered. When considering only 20% output coupled power from the two out coupled powers which are from 20% and 10% output couplers, we can achieve around 3.7% WPE.

72 Table 3 - 1. Wall-Plug Efficiency (WPE) of the Theta cavity and a SCOWA amplification

system.

Components Current (A) Resistance (Ω) Voltage(V) Power (W) Note

SOA (Covega) 0.6 *0.84 0.3024 TEC 0.26 **0.82 0.0554 Modulator 50 4.09 0.3346 RF Amplifier 0.17 15.48 2.6316 10dB coupling

Synthesizer (AC)*** 0.0002 (-6.9dBm)

SCOWA 3.5 1.31 4.585 TEC 0.997 **0.82 0.8151 * Series Resistance From Data Sheet SUM= 8.7243 Watt ** Melcor Web site. *** Direct RF Power. for 246mW Note: Thermisters are not included. WPE= 0.0281 Optical Power Two output couplers for 20% and 10%.

Another critical performance for mode-locked lasers is noise performance. An absolute

and a residual phase noise measurement system characterize the photo-detected signal noise of optical pulse train. However, before measuring the phase noise, an RF spectrum analyzer gives us an indicative noise performance even though the dynamic range of an RF spectrum analyzer is not enough for observing fine details of the phase noise. Fig. 3-27 shows the photo-detected RF signal from the short pulse of the Theta cavity after the SCOWA amplification. The super-mode suppression ratio of the 106MHz is around 98dBc/Hz. For further investigation, the single side band residual noise measurement is needed.

0 Res: 30KHz

-20

98dBc/Hz -40

-60

Spectral Power(dBm) -80

102 104 106 108 110 Frequency (MHz)

Fig. 3 - 27. RF spectrum of photo-detected signal.

3.6 <100MHz repetition rate operation of the Theta cavity

As mentioned earlier, the Theta cavity has the ability to reduce the repetition rate by

mitigating the nonlinear effects of the semiconductor gain medium in the cavity. Low repetition

rate operation of a semiconductor gain medium has the scalability of the pulse energy and

spectral broadening. Hence low repetition rate operation with a low group delay CFBG can show an ultimate advantage of the Theta Cavity for generating direct frequency-swept pulses with high power and broad bandwidth. Even though the CFBGs impose some limits on the performance of

74 the theta cavity system, the limitation of the Theta cavity presently, broad optical band width and

low repetition rate operation is summarized in this section.

(a) (b)

Fig. 3 - 28. 33MHz repetition rate operation of the Theta cavity with 660ps/nm CFBG. (a) Short

pulse on sampling scope, and (b) frequency-swept pulse on spectrally resolved streak

camera.

The CFBG is used in this experiment has a group delay of 660ps/nm and an optical band pass filter with 20nm is used. The output power of frequency-swept pulses is 2.6mW, which is limited mainly due to the unfilled time interval of the frequency-swept pulses. However, in this case the limitation for filling the repetition interval of the frequency-swept pulses is a 20 nm optical band pass filer. In Fig. 3-28 (b), the grating to measure the spectrally resolved streak camera is 50g/mm which is not enough to resolve the fine details. Hence the optical spectrum analyzer (OSA) is used to measure the optical spectrum of the frequency-swept pulses for detail structure.

75 25 -20

20 -25 Power(dBm) Spectral

15 -30 10 -35 5 -40 0 -45 -5 GDR(ps) -10 -50 Red port GDR -55 -15 Blue port GDR -20 Mode-locked Spectrum -60 1540 1550 1560 1570 Wavelength(nm)

Fig. 3 - 29. Group delay ripple (GDR) of CFBG and mode-locked spectrum.

Even though we broaden the mode-locked spectrum with respect to repetition rate of the mode-locked laser, the intrinsic effects of group delay ripple degrade the mode-locked quality. In

Fig. 3-29, we observe the modulation on the mode-locked spectrum is closely related with the group delay ripple (GDR) variation over the bandwidth. A further study will be needed to investigate the limitation of GDR in the Theta cavity.

A 40nm optical band pass filter is inserted to broaden the optical spectrum in spite of the

GDR effect. A group delay of 991ps/nm CFBG is used as a stretcher and compressor in the

Theta cavity.

-20

-40

-60

-80 Spectral Power(dBm)

1500 1520 1540 1560 1580 1600 Wavelength(nm)

(a) (b)

Fig. 3 -29. A repetition rate of 31MHz with a group delay of 991ps/nm CFBG. (a) 31MHz

sampling scope trace, and (b) its mode-locked spectrum.

Frequency swept pulses of 33.8nm (15dB) at 31MHz with a group delay of 991ps/nm

CFBG and 138pJ pulse energy were achieved. The 31MHz repetition rate which is ~32ns

repetition interval is fully filled with 33.8nm (15dB) stretched pulses.

In conclusion, the Theta cavity is operated at ~30MHz low repetition rate taking

advantages of extreme stretched in the semiconductor gain media. This low repetition rate operation enable to achieve more than 130pJ from frequency swept pulses and 33nm (15dB) mode-locked band width.

77 3.7 Preliminary experiment of feed forward intensity noise cancellation and gain

flattening scheme for frequency swept pulses

Low noise operation in terms of intensity and phase is crucial for application. In this

section novel scheme for gain flattening and intensity noise cancellation using feed forward

technique is introduced. The feed forward technique is generally better than the feed back [14]

technique in terms of bandwidth.

(a) (b)

Fig. 3 - 30. Feed forward scheme for frequency swept pulses. (a) General schematic, and (b)

concept.

Frequency swept pulses are divided into two using a beam splitter or an output coupler,

and one of them is photo-detected and the photo-detected signal is amplified and then finally this

signal match the originally optical signal in a intensity modulator. An optical signal from an

upper arm is equal to the path length for a RF signal which is a lower arm in Fig. 3-30 (a). This feed forward scheme for the frequency swept pulses has a gain flattening mechanism, additionally, due to the wavelength to time mapping in stretched pulse. The concept of gain

78 flattening and noise cancellation in the feed forward scheme for frequency swept pulses is introduced in Fig. 3-30 (b).

1.0 1.0

0.8 0.8

0.6 0.6

After Feed Forward 0.4 Before Feed Forward 0.4 (BW 500MHz-1000MHz) (BW 500MHz-1000MHz) Stretched Pulse w/ 511MHz Stretched Pulse w/ 511MHz 0.2 0.2 Spectral Power(A.U) Spectral Power(A.U) 0.0 0.0 1546 1548 1550 1552 1554 1546 1548 1550 1552 1554 Wavelength(nm) Wavelength(nm) (a) (b)

(c)

Fig. 3 - 31. Feed forward noise cancellation scheme. (a) Before feed forward scheme, (b) after

feed forward scheme, and (c) experimental setup.

79 A 500~1GHz band width RF amplifier is used to increase the power of the photo- detected signal. The theta cavity is operated at 611MHz with quasi CW mode. An 80: 20 split ratio output coupler is used. An 80% output beam is detected with 12GHz fast photo-detector.

The path length is tested using a short pulse. After the path length matching, the optical spectrum is monitored to investigate the gain flattening effect. A DC-biased voltage of intensity modulator is on the down slope of transmission curve to reverse the photo-detected signal.

The preliminary result of the feed forward scheme is shown in Fig. 3-31. It is clear that the modulation depth with respect to the spectrum is suppressed. Due to wavelength and time mapping, the decrease in the modulation depth implies that the intensity noise, which is a spectral fluctuation, is reduced. However, a further study is needed to quantify noise cancellation performance.

3.8 Summary

In this chapter, we have developed a novel mode-locked oscillator which is called eXtreme Chirped Pulse Oscillator (XCPO) based on the Theta cavity design. This oscillator has distinctive features.

z Energy scalability

z Optical band scalability

z Low repetition rate operation

z Directly frequency-swept pulses

z Excellent wall-plug efficiency with high power amplifiers

80 We have achieved more than 130pJ directly from the Theta cavity at 31MHz. The mode- locked bandwidth of about ~34nm (15dB) is also achieved. This direct stretched pulse generation from the oscillator can be used in many applications such as swept-source optical coherence tomography (OCT) [10], and time stretched photonic analog to digital converters (P-ADC) [11].

81 CHAPTER 4

GRATING COUPLED SURFACE EMITTING LASER AND GRATING

COUPLED SURFACE EMITTING SEMICONDUCTOR OPTICAL

AMPLIFIER

4.1 Introduction

Grating-coupled surface-emitting lasers (GCSELs) offer many unique advantages for picosecond optical pulse generation, such as a wide spectral tunability [1], large output transverse mode properties and high power based on a GCSEL MOPA system [2]. In addition, an external cavity incorporating a GCSEL can utilize the grating dispersion for spectral filtering, gain flattening [3] and wavelength tuning for a mode-locked laser.

In this section, we investigate the passive and hybrid mode-locking of a GCSEL and the energy extraction from a grating-coupled surface emitting semiconductor optical amplifier [4] using the XCPA concept [5].

Fig. 4 - 1. Grating-Coupled Surface-Emitting Laser.

The GCSEL is composed of one gain section and one passive section as shown in Fig. 4-

1. The passive section of the GCSEL in this experiment incorporating a grating coupler with an

HR coating serves two purposes in the external cavity: First, the grating couples out power confined in the waveguide to radiate out because the substrate has low reflectivity. Secondly, it acts as a saturable absorber that provides a mechanism for passive and hybrid mode-locking. In

83 edge-emitting laser diodes, the use of an unpumped section as a saturable absorber without proton bombardment has been reported [6].

4.2 Grating coupled surface emitting laser

The GCSEL device is fabricated from an AlGaAs graded-index separate-confinement

heterostructure (GRINSCH) with a single InGaAs quantum well. The p-doped cladding in the

passive region, where the grating coupler is fabricated, is wet-etched down to 250nm above the

GRINSCH, using an InGaP etch stop layer. The wafer structure is shown in Table. 4-1.

Table 4 - 1. GCSEL wafer structure.

Layer Composition Thickness(µm)

p-contact GaAs:C 0.2

p-cladding Al(y)Ga(1-y)As:C 0.75

Etch-stop InGaP:Zn 0.02

p-cladding Al(y)Ga(1-y)As:C 0.25

waveguide Al(x)Ga(1-x)As->Al(y)Ga(1-y) 0.2

Buffer layer GaAs 0.01

QW In(1-x)Ga(x)As 0.008

Buffer layer GaAs 0.01

waveguide Al(y)Ga(1-y)As->Al(x)Ga(1-x) 0.2

n-cladding Al(y)Ga(1-y)As:Si 1.2

Buffer layer GaAs:Si 0.2

Substrate n-GaAs Substrate 170

The p-contact of the 60 micron wide gain-guided device is evaporated and annealed on the active section. The 270nm period grating is patterned in resist using electron beam lithography and is then etched by an ICP etcher using chlorine-based chemistry. The corresponding diffraction angle in air is ~17 degrees. Upon inspection with a scanning electron microscope (SEM), the grating depth is measured to be 220nm for a duty cycle close to 35%. A

SEM picture of grating is shown in Fig 4-2.

InGaAs QW

Fig. 4 - 2. A scanning electron microscope (SEM) image of grating.

86 The n-type Ni/Ge/Au contact is patterned and aligned with the p-side active section by a mask aligner equipped with infrared viewers. When patterning the n-contact, a non-metallized window is created to allow light extraction through the GaAs substrate. A high reflection coating composed of a 130nm dielectric layer and 225nm of gold is deposited on the grating to enhance the directionality of the grating. A low reflection dielectric coating is deposited on the n-side

GaAs windows and is dry-etched on the n-side contact to enable proper wire bonding of the device. The GCSEL device, with 1.7mm-long active and 0.6mm-long grating sections, is cleaved from the processed wafer and is then bonded p-side down on a BeO heatsink with In solder.

For device performance evaluation, light-current and spectral characteristics are investigated under DC pumping. The substrate output slope efficiency is 0.53W/A as shown in

Fig 4-3 (a) and the lasing threshold is measured to be ~ 750mA. For lower DC pumping, no significant lasing behavior is observed. It should be noted that for measuring the spectrum of the grating port, a basic 4-f system is used to match the angular dispersion and image of the grating.

A 100µm-core diameter multimode fiber is used to couple all spectrum of the grating port. The spectrum difference between grating port and cleaved facet presented in Fig. 4-3 (b) and (c) results from the grating’s transmittance for wavelengths.

30 Power(mW)

0 0 50 100 150 200 250 300 Current(mA)

(a)

100mA 100mA 200mA 200mA 300mA 1E-3 300mA 1E-3

1E-4

1E-4 1E-5 Spectral Power(dB) Spectral Spectral Power(dB)

1E-6

1E-5 950 960 970 980 990 950 960 970 980 990 Wavelength(nm) Wavelength(nm)

(b) (c)

Fig. 4 - 3. The GCSEL performance (a) Light vs Current curve of grating port, (b) Spontaneous

emission spectrum of cleaved facet, (c) Spontaneous emission spectrum of grating port.

88 4.3 Passive and hybrid mode-locking of grating coupled surface emitting laser

4.3.1 Experimental setup for mode-locking

We demonstrated active mode-locking of an external cavity GCSEL [7]. In active mode- locking experiments, the grating port was used to provide optical feedback whereas the substrate port was used as an output for constructing the cavity. The slit parallel to the grating in the cavity was used to filter the spectral modulation resulting from the substrate reflection (R~30%).

However, the GCSEL used in this experiment has a high reflection coating on the grating to sum the total power through the substrate, and an anti-reflection coating on the substrate to suppress feedback. The role of the slit in the cavity, which is normal to the grating, controls the transverse mode profile. In this experiment, we demonstrate a passively and hybridly mode-locked GCSEL external cavity using its passive section as a saturable absorber.

The mode-locked laser setup is depicted in Fig. 4-4. We use 0.5NA lenses for collimating the output light from the GCSEL and for focusing light onto the HR mirror. The cleaved facet, having a reflectance of ~ 30%, is used as the output port of the laser. The spatial beam profile and polarization of the output beam in the external cavity are rotated to match the beam modes of an angle striped semiconductor optical amplifier to increase the signal.

Fig. 4 - 4. The experimental setup for passively and hybridly mode-locked external cavity using

GCSEL and sketch of GCSEL.

MLL-Mode-locked laser, SOA-Semiconductor optical amplifier, GCSEL-Grating coupled surface emitting laser, G-Grating, S-Slit, HR-High reflector, P-Pellicle, L1, L2-Lens, f1, f2-

Effective focal length, CCD-Charge-Coupled-Device, AC-Auto-correlator, SC-Optical sampling scope, and OSA-Optical spectrum analyzer, RFSA-RF spectrum analyzer.

90 In order to compensate the linear chirp of the generated mode-locked pulse, a dispersion compensator consisting of bulk gratings and a telescope is used. The output of the mode-locked laser is detected with a 25GHz fast photodiode and monitored with a 50GHz sampling scope and

RF spectrum analyzer. Optical autocorrelation techniques are used for measuring pulse width characteristics below the resolution of the electrical measurement system.

To increase the cavity stability of passive mode-locking, the cat’s eye geometry is used.

Owing to the wide stripe of the GCSEL device, the width of the intracavity slit, which is normal to the grating, is optimized to control the transverse mode profile which makes a small spot on the unpumped section in order to improve pulse quality in passive mode-locking. The slit in the

Fourier plane of the dispersion compensator is also adjusted to improve the spectral characteristics.

4.3.2 Passive mode-locking of GCSEL

The optimum passive mode-locking regime operates at a 535MHz repetition rate, which is the fundamental frequency of the cavity. This regime is observed at a DC current of 176mA.

The average optical output power of the oscillator is 4.2mW.

The output pulse is measured via intensity autocorrelation technique. The pulse chirp is then removed using the dispersion compensator generating pulses of 7ps in duration. (See Fig. 4-

5 (a)). In order to confirm passive mode-locked operation, the light vs current characteristics are measured with high resolution (Fig. 4-5 (b)). The hysteresis is an indication of passive mode- locking and shows the nonlinear bleaching process of the unpumped section. In typical CW

91 lasers without saturable absorption, a hysteresis in the light vs current characteristics is not

observed. Fig. 4-5 (b) (inset) shows the light vs current characteristics up to 200mA.

Additional evidence for passive mode-locking can be obtained by examining the near

field spatial mode characteristics of the output beam, since the near field image is a direct

measure of the spatial characteristics of the unpumped region. When measuring the near field

image, a 4-f optical system including the first collimating lens and a pellicle is used to match the dispersion of the grating. Even though multimode behavior is obtained from the wide gain strip

width, only a localized spot is influenced by the slit in the cavity, which drastically changes

corresponding to the generation of short pulses as shown in Fig. 4-5 (c) and (d).

92 1.0 uncompressed pulse FWHM 38.2 ps 4.5 compressed pulse 0.8 (a) (b) FWHM 8.8 ps 4.0

0.6 3.5

3.0 0.4 8

Power(mW) 2.5 4 0.2 Power(mW)

Autocorrelation Signal(A.U) 2.0 0 0 100 200 Current(mA) 0.0 1.5 -80-60-40-200 20406080 160 165 170 175 180 Delay(ps) current(mA)

Fig. 4 - 5. (a) Autocorrelation trace of compressed and uncompressed optical pulses in passive

mode-locking. (b) Hysteresis of passive mode-locking and LI curve of cavity (inset).

4.3.3 Hybrid mode-locking of GCSEL

In order to achieve hybrid mode-locking, the same optimal conditions are used with the addition of a 535MHz RF drive signal superimposed onto the gain section of the GCSEL using a bias tee. The best hybrid mode-locking which shows a 36.8psec uncompressed autocorrelation pulse duration occurred with 1W of RF power and a 153.6mA DC bias current. It should be noted that no effort is made to impedance match the RF signal to the GCSEL device.

The RF spectrum in Fig. 4-6. (a) shows the optimized hybrid mode-locking at 535MHz which is the fundamental frequency of the external cavity. The relatively high amplitude of the first harmonic frequency implies good performance of the hybrid mode-locking. A sharpening of the 535MHz tone states that the cavity lasing modes are well locked to the external RF drive signal. The inset shows the RF spectrum of the passive mode-locking. Fig. 4-6. (b) shows the

optical spectrum of the hybridly mode-locked oscillator measured after the dispersion

compensator. The optical bandwidth of the hybridly mode-locked laser after spectral filtering is

~0.6nm. The main limitations of the spectral broadening are the AR coating performance on the

substrate and the grating dispersion mismatch of the cavity.

94 -10 -10 RES : 100KHz RES:100KHz 1.0 -20 -20 (a) -30 -40 0.8 (b) -30 -50 -60 -40 -70 0.6 Spectral Power(dBm) Spectral -80 530 532 534 536 538 Frequency(MHz) -50 0.4 -60 Spectral Power(dBm)Spectral Spectral Power(A.U) 0.2 -70

-80 0.0 530 532 534 536 538 540 974 975 976 977 978 979 Frequency(MHz) Wavelength(nm)

1.0 uncompressed pulse ) FWHM 27.1 ps compressed pulse A.U ( 0.8 (c) FWHM 5.9 ps transform-limited FWHM 4.2 ps 0.6

0.4

0.2 Autocorrelation Signal

0.0 -60 -40 -20 0 20 40 60 Delay(ps)

Fig. 4 - 6. (a) Fundamental frequency RF spectrum measured with 100KHz resolution

bandwidth in hybrid mode-locking and passive mode-locking (inset). (b) Optical

spectrum in hybrid mode-locking. (c) Autocorrelation trace of compressed optical

pulses in hybrid mode.

95 The autocorrelations of uncompressed and compressed optical pulses in hybrid mode-

locking are shown in Fig. 4-6 (c). Fig. 4-6 (d) shows the sampling scope image of the

uncompressed hybrid mode-locked pulse train. Assuming a hyperbolic secant-squared pulse

profile, the compressed autocorrelation trace in hybrid mode-locking implies a FWHM pulse

duration of 3.4ps, which is 3 times the transform limit, and the average output power of the oscillator is 3.9mW, implying a peak power of 0.305W, with a repetition rate of 535MHz.

4.3.4 Hybrid mode-locking with dispersion matched cavity of GCSEL

The linear cavity which used in the previous experiment needs to match dispersion of cavity due to a grating structure in the cavity. In order to match dispersion, a cylindrical lens is inserted satisfying 4-f system with 3 spherical lenses shown in Fig. 4-7.

(a) (b)

Fig. 4 - 7. Dispersion matched cavity of GCSEL. (a) Experimental schematic, and (b) optical

layout with respect to vertical and horizontal directions.

S: Slit, Cyl: Cylindrical lens, HR: High reflector, GCSEL: Grating Coupled Surface Emitting

Laser, f: Focal length (f1=18.4mm, f2=150mm, f3=18.4mm).

The focal length of the cylindrical lens is 7.2mm. A DC current of 85mA with a 376MHz

RF signal is applied in order to achieve hybrid mode-locking. The output power of 2mW in the hybrid mode-locking is achieved with 376MHz repetition rate.

1.0 1.0 Compressed Pulse FWHM 7.0ps 0.8 0.8

0.6 0.6

0.4 0.4

Power(A.U) Spectral

Autocorrelation Signal(A.U) Autocorrelation 0.2 0.2

0.0 0.0 974 976 978 980 -60 -40 -20 0 20 40 60 Wavelength(nm) Delay(ps) (a) (b)

Fig. 4 - 8. Hybrid mode-locking performance. (a) Intensity auto-correlation, and (b) its mode-

locked spectrum.

In this experiment, an auto-correlation width of 7ps and a mode-locked bandwidth of

~0.6nm are achieved. The results from the experiment are similar to the previous experiment.

Therefore, it is obvious that the dispersion of the cavity is not a main limitation for a short pulse and a broad spectrum of mode-locking. However, generally, the dispersion matched cavity is a better solution to generate short pulses and broaden mode-locked spectra.

As a future study in order to enhance saturable absorption, a lateral injection device for

GCSEL might be a good candidate. A general lateral current injection scheme and a proposed

GCSEL structure for better mode-locking are shown in Fig. 4-9.

(a) (b)

Fig. 4 - 9. Proposed 2-section GCSEL. (a) Lateral current injection laser diode [8], and (b) 2-

section GCSEL based on current wafer structure.

Fig. 4-9 (a) shows the general lateral current injection laser while (b) shows the proposed

GCSEL with lateral current injection scheme. The basic idea is similar with 2-section monolithic semiconductor mode-locked lasers [9] except grating section in saturable absorber. Monolithic muti-section semiconductors for high repetition rate have been widely investigated for mode- locking [10]. The proposed 2-section GCSEL might be good for spectrum tuning and short pulse generation using gain flattening.

4.4 Grating coupled surface emitting semiconductor optical amplifier

4.4.1 Grating coupled surface emitting semiconductor optical amplifier

The key feature of Grating-Coupled surface-emitting Semiconductor Optical Amplifiers

(GCSOAs) is that they have two grating couplers instead of one as depicted in Fig. 4-10 [4].

Because both facets have grating structures, the amount of the back reflection is significantly

reduced compared to the GCSELs, hence GCSELs can be used as optical amplifiers in

conjunction with the functional advantages of GCSELs. The GCSOAs are attractive for high

power applications because of their potential to overcome the catastrophic optical damage

(COD) level of the edge emitting laser diodes. COD is a failure of facets due to high intensity

which leads to melting and recrystallization of the semiconductor material at the facets of the laser [11]. Therefore, COD can be considered a limiting factor for high power performance in semiconductor lasers. In order to increase the COD level, a broad area laser diode [12] or a large optical cavity (LOC) structure laser diode [13] has been researched. In the edge emitting laser diode case, COD level is around 18–18.5 MW/cm2 [14]. Thus the surface-emitting lasers, such as

Vertical-Cavity Surface-Emitting Lasers (VCSELs) [15] and GCSOAs, whose emitting area can

be scaled, can potentially lead to increase the COD levels.

100

Fig. 4 - 10. Typical scheme of Grating-Coupled Surface-Emitting Semiconductor Optical

Amplifier.

4.4.2 GCSOA and CW amplification

As stated before, the XCPA system in conjunction with a GCSOA as a final amplifier can be an attractive choice for a high power laser system. This is because of the promise of the

XCPA system to extract more pulse energy for a given pumping level as stated in the previous chapter and the GCSOAs to increase COD levels compared to the edge emitting devices.

Before performing the XCPA system with GCSOAs, the CW injection performance for

GCSOAs is tested to observe its feasibility for high power applications. The XCPA system has a similar behavior to the CW operation due to the extreme stretching of pulses [5]. In this experiment, the GCSEL has 3 output ports including the cleaved facet used for monitoring the spectral properties and the substrate port used for as the output coupler.

101

400 OSC : CW 350 GCSOA : 1KHz/500nsec Date : OCT/2002 300

250

GCSO 200

150 GCSEL A 100 Peak Power(mW)Peak

50 Peak Current 4A f f f f Peak Current 2A 0 24681012 Injected Power(mW) (a) (b) Injected Spectra 1E-4 Spontaneous Emission 1KHz/500nsec/4A/12mW 1E-5 Oct/2002

1E-5

1E-6

Spectral Power(dBm) 1E-6

Power(dBm) Spectral 1E-7

960 965 970 975 980 960 965 970 975 980 Wavelength(nm) Wavelength(nm) (c) (d)

Fig. 4 - 11. CW injection to GCSOA driven by QCW (a) experimental setup, (b) amplified peak-

power vs. input power, (c) spectrum of the oscillator, and (d) injected spectrum vs.

amplified spectrum.

The GCSEL oscillator performing in CW operation is shown in Fig. 4-11 (a). The

GCSEL has a 1940um long gain section with a 60um wide stripe while the length of the grating

section is 600um. The GCSOA has a 4000um long gain section and a 330um grating length. In

order to match the angular dispersion of the GCSOA, the same periodicity for the grating, which

is 325nm, is fabricated. The GCSOA is pumped by a QCW electrical signal with 1 KHz repetition rate and 500 ns pulses to avoid the thermal loading of the device. In order to match the

102 angular dispersion and maximize the coupling, a 4-f optical system with an isolator is employed after the external cavity.

The CW laser spectrum is shown in Fig 4-11 (c) and the amplified spectrum, which represents the S/N of the amplification, is shown in Fig. 4-11 (d). At 4A peak current, the maximum peak power is 375mW from the substrate port of the GCSOA while 12mW of input power from the GCSEL external cavity is injected as shown in Fig. 4-11 (d). Assuming unity

S/N in this data, 0.187 uJ can be extracted from one facet at 500 ns electrical pumping. This result shows promise for the XCPA approach using GCSOAs as stretched-pulse amplifiers.

4.4.3 GCSOA and extreme chirped pulse amplification for energy extraction

Next experiment applies the amplification for the combination of the XCPA and a

GCSOA. An experimental setup of the combination of XCPA and GCSOA amplification is depicted in Fig. 4-12 (a).

103 CPSML

G G Pulse Picker

Pre-amplifier Block (Tapered SOA + CFBG Single mode SOA) (20ns stretched pulse) Amplifier Block (GCSEL+ GCSOA)

(a)

-30 1.0 -10 FWHM 1.35 ps -20 0.8 -40 -30 0.6 -50 -40 AC 0.4 -50 Power(dBm)

Power(dBm) -60 0.2 -60 Hyperbolic Secant: 0.875ps 0.0 -70 970 972 974 976 978 02468 960 970 980 990 Wavelength(nm) Delay(ps) Wavelength(nm)

(b) (c) (d)

Fig. 4 - 12. Energy extraction from GCSOA setup and general performance by using XCPA. (a)

experimental setup, (b) spectrum of CPSML, (c) autocorrelation of CPSML , and (d)

amplified spectrum of GCSOA.

104

In this experiment, the previously developed CPSML is used as a master oscillator and its

performance is summarized in Fig. 4-12 (b) and (c). The repetition rate is 285MHz and the pulse

width 875fs. A 4000ps/nm CFBG made by 3M is used as a stretcher, and a pulse picker

employed by an SOA is electrically pulse-biased to avoid the pulse overlap due to the stretcher.

Hence, the repetition rate goes down to 47.5MHz because the bandwidth of the CFBG and the

CPSML is around 5nm. After the stretcher, a pre-amplifier chain, which is composed of a single

mode SOA and a tapered SOA, is used to boost the signal power.

At final stage of amplification, a GCSOA is inserted to improve the coupling efficiency,

however, because the input of the device is a grating coupler, spatial mode matching is required.

A GCSOA is employed to match the spatial mode of the final amplifier. The parameters of the

GCSEL and GCSOA used in the experiment are summarized in Table 4-2.

105 Table 4 - 2. The parameters for GCSEL and GCSOA which are used in this experiment.

GCSEL GCSOA

Chip# S19SDSS#3-22b S19SDSS-#2-10

Wafer# A2170-57 A2170-57

grating periodicity (nm) 270 270

grating depth (nm) 245-255 245-255

active length (µm) 2000 4000

grating length (µm) 750 750

stripe width (µm) 100 200

Threshold current (mA) 219.2 879

Total slope eff (W/A) 0.50 N.A

Max conversion eff (%) 21% N.A

Serial Resistance (Ω) 0.12 N.A

Between the GCSEL and the GCSOA, a traditional 4-f optical system is changed to an 8- f system as shown in Fig. 4-13 (a) because of the clear aperture’s size of an isolator without losing an angular dispersion matching. The 8-f system is composed of 18.4mm and 75.0mm of effective focal length lenses.

106

Input GCSEL

f1 f1

f1 Output

f1 GCSOA

(a) (b) (c)

Fig. 4 - 13. Mode-matched optical system (a) setup, (b) output spatial mode before injection and

To investigate the spatial mode of the GCSOA, a laser beam analyzer from Spricon is placed on the output of the GCSOA. As depicted in Fig. 4-13 (b) and (c), the longitudinal direction is the grating groove direction analyzing the wavelength distribution along the longitudinal direction while the vertical direction is related to the device’s width. The comparison between the two spatial modes roughly indicates that the spectral power of the mode-locked laser is fully injected.

The input average power before the GCSEL is 300mW. The GCSEL as a spatial mode converter, is biased with 200mA DC current and the GCSOA is pumped by the electrical QCW

−5 with a 250nsec pulse duration and a 3.75× 10 duty cycle in order to avoid the thermal load.

The performance of the combination of XCPA and GCSOA is summarized in Fig 4-14.

At 100A, we get 24.5W of peak power which is almost 0.5µJ from the 20 ns stretched pulses after considering the optical signal to noise ratio as shown in Fig. 4-14 (a). In Fig. 4-14 (b), as

107 the current is increased, the spectral power outside the mode-locked band is increased. This is mainly due to the substrate reflection. One can prove that the periodicity of the ripple is exactly matched with the substrate thickness. This results in degradation of the optical signal to noise ratio and subsequently the energy per pulse.

-30 25 10A 40A 24.5W 70A 20 -40 100A

15 -50

10 Signal -60

Power(dBm) Band

Peak Power(W) 5 -70 0 0 20406080100 960 970 980 990 Peak Current(A) Wavelength(nm)

(a) (b) 20

S/N

0 0 20406080100 Peak Current(A)

(c)

Fig. 4 - 14. XCPA and GCSOA performance. (a) Peak powers for 20nsec stretched pulse, (b)

spectral evolution as increasing biased current, and (c) optical signal to noise ratio.

108 To suppress the feedback from the surface, a high quality AR coating is required on the substrates. However, it should be noted that the angular dispersion from the grating in terms of wavelength makes a different incident angle with the AR coatings on the substrate.

The thermal rollover, which limits energy extraction from the GCSOA, is observed at high currents, i.e, 100 A. When the current was 100A, the failure of the device was observed. It is expected that degradations other than the COD may limit the performance because both the grating couplers mitigate the COD effect due to their broad emitting areas. In order to investigate the electrical break down of the PN junction of the quantum well for the failure, the IV curve of the GCSOA was measured again. No difference between the new and old IV curves was observed. We suspect the grating surfaces, which are coated by SiN/Au, may be damaged due to the difference in thermal expansion. In order to overcome this effect, a further study will be required.

4.5 Summary

In conclusion, the passive and hybrid mode-locking of an external cavity grating-coupled surface-emitting laser by using an unpumped section as a saturable absorber has been successfully demonstrated. In hybrid mode-locking, pulse widths of 3.4ps were achieved. We anticipate further reduction in the output pulse width by improving AR coating performance and using a dispersion matched cavity.

In order to improve the mode-locking performance, we have proposed a new lateral current injection GCSEL with an active saturable absorber.

109 It was experimentally demonstrated that GCSOAs are good candidates for high power laser amplification in the XCPA regime. Around 0.5µJ pulse energy for 20ns stretched pulses was obtained.

110 CHAPTER 5

PASSIVE OPTICAL CAVITY AMPLIFICATION

5.1 Introduction of passive optical cavity

A passive optical cavity (POC), or an enhancement cavity, consisting of polarization

components such as a polarizer, a birefringent crystal or a Brewster plate having polarization-

dependent losses in the cavity is an amplifier system to store energy of optical pulses or CW

laser. The passive optical cavity amplification of mode-locked pulses is easily understood as a

coherent superposition of electric field of mode-locked pulses in time domain and an optical

frequency-comb enhancement of mode-locked spectra as shown in Fig.5-1 [1]. The stored energy

in the closed passive cavity is finally switched out using an acoustic optic modulator (AOM) or a

dumping device. More importantly, a special situation when the reflectivity of the input coupler

in the POC exactly equals the additional losses of the cavity causes the net reflection of the cavity to become zero. This has been called the impedance matching [2].

The finesse of a cavity is related to the maximum energy storage capability and this can

be written as

c F*2= πτ 2l [5.1]

where, F * is the finesse, c is the speed of light, l is the length of a cavity and τ is the decay

time of the cavity. Based on the above equation, we can estimate the dumping rate or

amplification factor for a given finesse and repetition rate. Hence,

111 F * M ∝ [5.2] 2π can be used as a figure of merit for amplification of a POC. Based on both characteristics of the impedance matched conditions and the finesse of the cavity, the maximum amplification is theoretically limited by the finesse of the cavity when its impedance is matched.

Reduced Repetition Rate

Dumped Output

Cavity Dumper

High Q-cavity (a) (b)

Fig. 5 - 1. Passive optical cavity. (a) Time domain [1], and (b) frequency domain explanation.

5.2 Hänsch-Couillaud stabilization

Another consideration of the POC is the stabilization scheme based on its needs of interferometric stabilization to accumulate the frequency combs of a mode-locked laser. In our case, the polarization-sensitive Hänsch-Couillaud stabilization scheme [3] is used for this stabilization. According to the Hänsch-Couillaud stabilization scheme, the reflection of the light from the input coupler can be a reference in terms of phase. If the cavity length differs from the resonance frequency of a POC, the output light from a POC has an accumulated phase difference

112 with respect to the reference beam. The combined beam from the reflection of the input coupler and the POC results in an elliptical polarization due to the phase shift when the cavity length is slightly changed. The analyzer part for this phase shift consists of a Quarter Wave Plate (QWP), a Polarization Beam Splitter (PBS) and a balanced detector which leads to an error signal that has the phase difference information for a locking system having a Proportional-Integral-

Derivative (PID) controller and a piezo-electric actuator. Basically, the Hänsch-Couillaud is similar as the Pound-Drever-Hall (PDH) stabilization method [4], since both approaches employ the fact that if the cavity is off resonance from the ideal value, a phase shift is generated after round trips in the cavity indicating a change in the cavity length. The comparison between both schemes is summarized table 5-1.

These active stabilization schemes of reference cavity are widely used for efficient second or higher harmonic generation [5], mode-locked laser stabilization [6] in the cavity. A passive optical cavity using a commercially available picosecond mode-locked laser has been demonstrated by Potma et al [7] using Hänsch-Couillaud stabilization. In this pulse width regime, dispersion in the cavity is not a crucial concern for high peak power generation, however, in a few femtosecond pulses, dispersion in the passive optical cavity should be considered [1, 8].

More promising application for passive optical cavity is intracavity extreme ultra-violet (EUV) generation [9].

113

Table 5 - 1. Comparison between Hänsch-Couillaud method and PDH method [10].

Hänsch-Couillaud Pound-Drever-Hall

Polarization sensitive stabilization Concept RF sideband modulation technique of resonator length Polarization of fundamental beam is rotated slightly Fundamental laser is phase modulated before cavity Bireferingent material Beat note between transmitted and or Brewster plate reflected light is measured Principle of Polarization of exit beam is Phase of beat signal is compared operation analyzed (QWP+PBS) with oscillator phase On Resonance: linear polarization On Resonance: phase of beat signal (PD1=PD2) =oscillator phase Off Resonance: elliptical Off Resonance: phase of beat signal polarization (PD1≠PD2) ≠oscillator phase Adjustable locking range, defined by Easier to implement modulation frequency Pro Sideband modulation easily realized with No laser modulation required diode lasers Temperature and alignment Con EOM/AOM needed for non-diode lasers sensitive

5.3 Experimental setup and results

First of all, in order to build a passive optical cavity, the length of cavity which is usually same length of a master oscillator is considered. The cavity length matching between both cavities can be determined using a spatial interference fringe from the passive optical cavity. The repetition rate of the POC is 95 MHz which is one third of the mode-locked laser. This means

114 that the every three of the optical combs from the CPML laser can be enhanced because the master laser, i.e., the colliding pulse semiconductor mode-locked laser (CPSML) which has developed before, has 285MHz repetition rate in this experiment. The length of passive optical cavity is determined by observing the spatial interference fringe from a pellicle, which is just for monitoring, in the cavity. In this case, the resolution for matching length between both cavities is about pulse width, which is less than 2mm.

The TeO2 based AOM is aligned with the Brewster angle, which is 65.85 degree, for a minimum loss for TM polarization. The rotation of wedge type optics generates aberrations because the horizontal axis and the vertical axis of light beam have different incidence angle when the beam is convergent and also curved folding mirror generates astigmatism due to effectively different radius of curvature for the horizontal and vertical axis. However, the combination of the curved folding mirror resonator and the internal Brewster-angle cell can compensate the astigmatism which is generated by each of them [11] since the polarities of astigmatism generated from the AOM and the curved mirrors are opposite to each other. Hence the Brewster angle and the rotational angle of the folding mirrors are important for making an astigmatism compensated POC.

A complete setup for passive optical cavity dumping is shown in Fig 5-2. A Passive optical cavity (POC) used in this experiment is highlighted in the rectangular box in Fig 5-2.

115

Fig. 5 - 2. Passive optical cavity schematic.

CPSML: Colliding pulse semiconductor mode-locked laser, comb: Electrical comb generator,

QWP: Quarter wave plate, PBS: Polarization beam splitter, PD: photo-detector, PID:

Proportional-integral-derivative controller, HV amp: High voltage amplifier, PZT: Piezoelectric actuator, AOM: Acousto-optics modulator, M: mirror, HR: Super-polished high reflector.

116

Curved AOM Curved Mirror Mirror

Mirror

Mirror Input Coupler

Fig. 5 - 3. Picture of passive optical cavity.

By using the ASAP software program for simulation, 7.5 degree rotation of the folding mirrors can compensate the astigmatism for the TeO2 based AOM’s Brewster angle, where the radius of the curvature of the two curved mirrors near the AOM is 10cm and others are flat mirrors. Additionally, two folding mirrors are inserted to make square shape.

Fig. 5-3 depicts the real setup of a passive cavity on the optical table. A pellicle is inserted to monitor a beam profile in conjunction with a charge coupled detector (CCD) and a constructive interference pattern using a photo-detector.

117

Fig. 5 - 4. Passive optical cavity (POC) design and beam profiles. The design beam profiles at 2

different positions (near AOM) of and a measured beam profile at one position.

In order to increase stability of the cavity, the POC has two focal points in the cavity shown in Fig. 5-4. One of the waists is located in the middle of the input coupler and the next mirror where the beam waist is about 1mm in the design, while the other is located in the AOM.

To accumulate pulse energy in the POC, a mode-matched optics outside the cavity is crucial because it play an important role in connecting the two different cavities. The mode matching optics enables to match incident beam to the POC and beam after round trips in the POC in order to conform the self-consistent property. After measuring the divergence angle and the beam waist of the fiber-coupled CPSML, a Keplerian type telescope is designed using Zemax, to match the beam profile in the POC.

118 The error signal coming from the POC is measured by scanning the piezo-electric actuator as shown in Fig. 5.5. In this case the input coupler has a 99% reflectance, while other mirrors including two curved mirrors, which are super-polished to minimize the scattering losses, have 99.998% reflectance. The steep slope at the resonance of the error signal can be used to lock the POC to the CPSML by properly adjusting values of PID controller in conjunction with a high voltage amplifier.

Voltage(A.U) -1

-2

-3 -0.010 -0.005 0.000 0.005 0.010 Time(sec)

Fig. 5 - 5. The Fabry-Perot resonance, the error signal and the scanning voltage of the POC with

scanning Piezo-electric actuator.

119 A dumping system is needed for extracting energy from the POC. The dumping system is comprised of a TeO2 based AOM, a mixer, a synthesizer and an arbitrary signal generator. The synthesizer generates a carrier frequency which is suitable for the AOM bandwidth while the arbitrary signal generator turns on and off for a desirable interval to dumped stored energy. In our system, the carrier frequency of the AOM is 400MHz and the temporal gating pulse width from the arbitrary signal generator is 8ns with a variable repetition rate. The systematic diagram and the performance of the dumping electronics are shown in Fig. 5-6. One of design consideration for the electronics is that GND is important outside of the temporal gating pulses, otherwise it adds another loss for the POC.

0.3 0.2 Comb Mixer 0.1

Volt 0.0

Signal -0.1 Generator -0.2

-0.3 -3.00E-008 0.00E+000 3.00E-008 time(sec)

Fig. 5 - 6. The dumping electronics before a high power amplifier.

The signal is injected into a broad-band high power RF amplifier to get enough power for operating the AOM. The TeO2-based AOM has the highest efficiency at 980nm, which is 20% diffraction efficiency in CW laser at 1W of RF power. In order to increase the AOM diffraction efficiency, 10W of RF peak power boosted by a broad band amplifier is injected into the AOM

120 from 1MHz to 50 KHz. At 50 KHz dumping rate, the pulse energy is increased by a factor 7.2 as shown in Fig 5-7.

8 8

6 6

4 4 Factor Factor 2 2 Amplification Amplification

0 0 0 200 400 600 800 1000 0 5 10 15 20 Dumped Frequency(KHz) Time(us)

Fig. 5 - 7. Dumped energy from the POC with respect to the dumping rate and dumping interval.

To analyze this result, we should think about the decay time of the POC which is related with the build-up time of the cavity as shown in Fig. 5-8. The decay time of the POC with a 99%

R input coupler is estimated to be ~ 0.5usec which can be converted into 2MHz. Hence, the stored intra-cavity energy depending on the finesse of the cavity should be saturated at around

~1-2 MHz range and the dumped energy should show the saturated behavior. However, the result shows a linearly dependent increase with respect to the dumping rate up to 20usec. Hence it may require finesse measurement for more careful analysis.

121

Fig. 5 - 8. Photon life time of the POC.

The dumped pulse as well as the intra-cavity pulses which are the superposition of the pulses in the POC is also important because it emphasizes the stabilization of the POC with respect to the CPSML. The autocorrelation of the dumped pulse is measured using an auto- correlator at 1MHz, which is the minimum measurable rate for our auto-correlator, as shown in

Fig. 5-9 (a). This is a compressed autocorrelation because the pulses from the CPSML are compressible. The compressed autocorrelation of pulses directly from the oscillator is shown in

Fig. 5-9 (b).

122 1.0 Pulse Width: 0.88 ps 31 ps 0.8

0.6 AC

AC(a.u) 0.4

0.2

0.0 02468 Delay(31ps/div) Delay(ps)

(a) (b)

Fig. 5 - 9. Autocorrelations (a) from the passive optical cavity dumping and (b) from the

oscillator.

From the above result, the locking ability of the stabilization scheme and the master oscillator stability has to be carefully considered. The autocorrelation depicted in Fig. 5-9 (a) represents that the locking bandwidth is not enough for the mode-locked pulse train.

Considerations for the bandwidth are that of the balanced receiver, frequency response of PZT with optics, and the CPSML stability. A passive cavity with 95MHz free spectral range (FSR) and 99% reflection for an in-out coupler has a 304 KHz line-width, however, the measured line- width of the CPSML laser was about 190 KHz using the self delayed heterodyne technique [12].

Hence the balanced receiver bandwidth, which is about 20 KHz, should be improved in order to have a constructive interference in the POC. In addition, to follow the master oscillator, the frequency response of PZT with optics should be investigated carefully.

123 5.3 Summary

In conclusion, a astigmatism-free passive optical cavity (POC) has been designed and demonstrated. The colliding pulse semiconductor mode-locked laser with 285MHz repetition rate is used as a master oscillator. The POC with 95MHz cavity fundamental is stabilized to the master oscillator using Hänsch-Couillaud stabilization scheme.

Even though there is still room for improvement in terms of performance of the POC, we are able to obtain a stable error signal which was transferred to a locking system properly and an increase in energy per optical pulse by a factor of 7 was achieved by using an all semi-conductor mode-locked laser.

In order to improve the performance of the POC, improvement of the frequency response of stabilization circuits including the balanced receiver and a PZT with mounts and a mode- locked laser with a fundamental repetition rate are recommended in order to suppress super- modes.

124 CHAPTER 6

SUMMARY

6.1 Finished work

A colliding pulse semiconductor mode-locked laser (CPSML) has been developed to be used as a master oscillator for eXtreme Chirped Pulse Amplification. A pulse width of 0.88ps with 5nm mode-locked bandwidth centered at 975nm is achieved. Other colliding pulse semiconductor mode-locked oscillators have also been reviewed and proposed. Using the

CPSML, K. Kim et al have demonstrated eXtreme Chirped Pulse Amplification [1]. The new master oscillator and power amplification system has generated record peak power from an “all- semiconductor mode-locked laser”.

An eXtreme Chirped Pulse Oscillator (XCPO) implemented with a Theta cavity and based on a semiconductor optical amplifier (SOA) has been presented for generating >10ns frequency-swept pulses and 3.6ps compressed pulses directly from the oscillator [2, 3]. In this experiment, we have shown the two distinct characteristics of the XCPO which are the scalability of the output energy and the mode-locked spectrum. By using these characteristics, we have obtained a pulse energy of 138pJ from the stretched pulse and a mode-locked optical bandwidth of 33.8nm (15dB) directly from the oscillator. The laser cavity design also allows for low repetition rate operation <100MHz, as well. The cavity significantly, reduces nonlinear carrier dynamics, integrated self phase modulation (SPM), and fast gain recovery in an SOA.

125 We have reported passively and hybridly mode-locked grating-coupled surface-emitting laser (GCSEL) using its umpumped section of the GCSEL as a saturable absorber [4]. We have obtained 8.8ps FWHM autocorrelation pulses in passive mode-locked operation and 5.9ps

FWHM autocorrelation pulses in hybrid mode-locked operation, which are the shortest pulses from a GCSEL external cavity. With hybrid mode-locking, a peak power of 0.31W and a spectral bandwidth of 0.6nm have been obtained. In experiment of XCPA with Grating Coupled Surface

Emitting Semiconductor Optical Amplifiers (GCSOAs), around 0.5uJ pulse energy for 20ns stretched pulses has been obtained.

Passive Optical Cavity (POC) amplification or an enhancement cavity experiment has been investigated. A stable error signal and stabilization using Hansch-Couillaud method have been achieved. The stable error signal was transferred to a locking system, properly, and an increase in energy per optical pulse by a factor of 7 was achieved by using a CPSML as a master oscillator.

6.2 Future direction

Each of the above achievements has a different application area depending on its characteristics. For example, the passive optical cavity has intensively been investigated for

XUV generation for intra-cavity applications using high power femtosecond lasers by Ye’s group [5] and Hansch’s group [6] and the GCSELs and GCSOAs aim for a high power application for mode-locking, gain-switching and CW operations.

126 The possibility for high power operation and broad band generation of XCPO and XCPA enables us to focus on a Photonics Analog to Digital Converter and Optical Coherence

Tomography. The XCPA and XCPO systems can be used for novel approaches in Photonic

Analog to Digital Conversion (P-ADC) [7] by using their frequency-swept characteristics.

The P-ADC was first demonstrated in 1979 by H. Tayler et al [8]. He used several

different- Vπ intensity modulators to convert analog signals to digital. Since the first demonstration of P-ADC was done by Tayler, many researchers have paid attention to possibilities of replacing and improving the electronic ADC (E-ADC) [9] with photonics technology. However, there are no commercial products for real applications so far. In spite of difficulties to make a commercial product, many photonics researchers have tried to solve the bottle neck of E-ADC by using the Time Domain Multiplexing (TDM), the Wavelength Domain

Multiplexing (WDM) [10] and the time stretch systems [11] as shown in Fig 6-1.

127

(a) (b)

(c)

Fig. 6 - 1. The P-ADC configurations. (a) TDM [10], (b) WDM [10], and (c) time stretch system

[11].

Our XCPA and XCPO may be a master oscillator design for the time stretched P-ADC system because our systems have directly stretched pulse trains with a swept frequency.

However, in order to realize a master oscillator, other parameters such as, timing jitter (pulse to pulse) and intensity noise of stretched pulses should be considered as a future direction. As previously stated, the gain saturation effect of SOAs and the feed forward noise cancellation scheme can help reduce the intensity noise. In terms of phase noise, the XCPA approach is more promising due to accumulated group delay ripple of CFBG in XCPO.

128 As an OCT master oscillator, there is a room for broad bandwidth of XCPO. Even though we have achieved 33.8nm for 31MHz, this bandwidth is not enough for high resolution OCT. In terms of high speed of frequency-swept source, the XCPO approach is better than the so called

Fourier domain mode-locking [12]. However, the group delay ripple and stitching in the fabrication of CFBGs is still a problem in the realization for broad bandwidth XCPO.

Another application of XCPO is a master oscillator for Chirped Pulse Amplification

(CPA) [13]. Due to slow carrier life time of EDFA, the high peak power and high energy in short pulses is successfully generated in the CPA approach. The combination of XCPO and CPA system is ideal for generating high peak power and high energy pulse in picosecond pulses. In this case, we have to use a free space dispersion compensator after amplification of direct stretched pulses to avoid nonlinearities of CFBG when CFBG is used as compressor.

Finally, the feed forward noise cancellation scheme in conjunction with the gain saturation effect of SOAs can be investigated further in terms of relative intensity noise.

129 APPENDIX A: TRANSFORM-LIMITED PULSE CALCULATION

130

Mode-locked Spectrum E(t), I(t) (From OSA) Inverse Pulse I(w) Fourier Transform

Auto-correlation

A.C & Deconvolution Factor

Fig. A - 1. Diagram for transform-limited pulse and autocorrelation.

[Source Code in MatlabTM] %From Modelcoked Spectrum----->Transform-limited Pulse and Autocorrelation %Written By Shinwook Lee clear all; close all; C=3*10^8; %Speed of Light N=501; %Sampling Point=Data points sp=load('Theta-2.txt'); %Angular frequency spectrum, please use your source file temp=sp(:,1); %Temporal Storage A(:,j)is the j-th column of A w=(2*pi*C*10^9)./temp; %Angular Frequency for nano-meter based sp_f=sqrt(sp(:,2)); %E(w)=Sqrt{I(W)} ws=w(251)-w(252); %Minimum sampling frequency [rad/sec] ts=2*pi/(ws*N); %Minimum Resolution in Temporal Domain ~500fs T=ifft(sp_f); %Inverse Fourier_Transf{E(w)} Tc=ifftshift(T); %Center frequency shift option

Ic=Tc.*conj(Tc); %I(t)=|E(t)|^2 SHG_ac=xcorr(Ic,Ic); %Autocorrelation

131 figure(1); tscale=-(N-1)*ts:ts:(N-1)*ts; %Time axis of Autocorrelation S=linspace(1,50,1); %Narrowing the time window for view point S=SHG_ac(N-24:N+25)/SHG_ac(N); %normalization for Autocorrelation I=linspace(1,50,1); I=Ic(252-24:252+25)/Ic(252); E=linspace(1,50,1); E=Tc(252-24:252+25)/Tc(252); t1=linspace(1,50,1); t1=tscale(N-24:N+25); t1=t1'; save('auto.mat','t1','S'); subplot(2,2,1); plot(t1,S) subplot(2,2,3); plot(t1,I) subplot(2,2,4); plot(t1,abs(E))

%%% Check FWHM in Time domain of Autocorrelation%%%%%%%%%%%%%%%%% tw1=intrplat(t1,S,1.973*10^-12); tw2=intrplat(t1,S,-1.973*10^-12); %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%%% Check FWHM in Time domain of Intensity%%%%%%%%%%%%%%%%%%%%% twI1=intrplat(t1,I,0.98*10^-12); twI2=intrplat(t1,I,-0.98*10^-12); %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

132 APPENDIX B: DESIGN CODE FOR PASSIVE OPTICAL CAVITY USING

ASAPTM

133

!!95MHZ CAVITY GAUSSIAN BEAM PROPAGATION Written By Shinwook Lee

!!FOR 2 FOCAL POINTS

SYSTEM NEW RESET

XMEMORY FULL

UNITS MM

!!CONSTANTS START

Z1=-40 !! BEAM WAIST CALCULATION START POINT Z2=40 !!BEAM WAIST CALCULATION END POINT MIRROR_ROTATION_ANGLE=6.5 !! DEGREE M1_Z_OFFSET=1.3 !!SMALL - DOWNWARD RASIUS_CURVATURE=100 DETECTOR_LOCATION=13 GAUSSIAN_ST_PT=-452 DIA=35 !! DIAMETER GAUSSIAN_BEAM_WAIST=0.999999 !! GAUSSIAN BEAM WAIST from 0.07 GAUSSIAN_BEAM_DIV_ANG=(980*10^(-9))/(4*GAUSSIAN_BEAM_WAIST*0.001) !! 0.1 !!WAIST

NBR=15

!! MAIN VARIABLE FOR ALIGMENT

A=693.9 !! DISTANCE FROM ST. PT TO 1ST MIRROR M3_Z_OFFSET=-1.4 !!SMALL

!! END: MAIN VARIBALE

B=A/(COS[MIRROR_ROTATION_ANGLE*2]) !!DISTANCE FROM 1ST. M TO !!INTERSECTION

C=(745.3739-A)/(COS[MIRROR_ROTATION_ANGLE*2]) !! DIS FR INTERSECTION TO !!2ND M

D=745.3739-A !! DIS FR 2ND M TO END PT

134 H1=A*TAN[MIRROR_ROTATION_ANGLE*2] !! HEIGHT FROM ST. PT TO !! INTERSECTION

H2=(745.3739-A)*TAN[MIRROR_ROTATION_ANGLE*2] !!HEIGHT FROM !!INTERSECTION TO END

AOM_LOCATION=D-52 !!AOM START COOR

ANG_AOM=-55.4051!!-55.4051 !!AOM_BREWSTER_ANGLE @1.45 & 980NM

AOM_OFFSET=3/(COS[ASIN[SIN[55.4051]/1.45]])*SIN[-ANG_AOM- 34.5895] !!AOM_THICKNESS

OP_PATH_LENGTH_DIFF=0!!5.2839-3 !!OPLD

Z_OFFSET=(AOM_OFFSET)/(SIN[2*MIRROR_ROTATION_ANGLE]) !!Z_OFFSET !!ALONG Z-DIRECTION (0 0 Z) !!CONSTANTS END

!! Define coatings

COATING PROPERTIES 0 1 'TRANSMIT' 1 0 'REFLECT' 0 0 'ABSORB'

!! Define Media

MEDIA 1.5 'BK7' 1.45 'SIO2' !! GEOMETRY DEFINITIONS BEGIN HERE

!!START: 1ST MIRROR SURFACE OPTICAL Z (A+M1_Z_OFFSET) 0 ELLIPSE 2@(DIA/2) OBJECT '1ST_MIRROR' INTERFACE COATING REFLECT AIR BK7

GROUP -1 ROTATE X (MIRROR_ROTATION_ANGLE) 0 (A+M1_Z_OFFSET) COLOR 3 !!END:1ST MIRROR

!!START: 2ND MIRROR

135 SURFACE OPTICAL Z (-D) (RASIUS_CURVATURE) ELLIPSE 2@(DIA/2) OBJECT '2ND_MIRROR' INTERFACE COATING REFLECT AIR BK7

GROUP -1 SHIFT Y (H1+H2) ROTATE X (MIRROR_ROTATION_ANGLE) (H1+H2) (-D) COLOR 3 !!END:2ND MIRROR

!!!!AOM START

SURFACE OPTICAL Z (AOM_LOCATION) 0 RECTANGLE [email protected] OBJECT 'AOM.FRONT' INTERFACE COATING TRANSMIT AIR SIO2

SURFACE OPTICAL Z (AOM_LOCATION+3) 0 ELLIPSE [email protected] OBJECT 'AOM.BACK' INTERFACE COATING TRANSMIT AIR SIO2

SURFACE TUBE Z (AOM_LOCATION) [email protected] (AOM_LOCATION+3) [email protected] OBJECT 'AOM.TUBE1' INTERFACE COATING ABSORB AIR SIO2 BOUNDS +.3 -.2

GROUP -3 SHIFT 0 (H1+H2) 0 ROTATE X (ANG_AOM) (H1+H2) (AOM_LOCATION)

COLOR 2

!!END OF AOM

!!START: 3RD MIRROR SURFACE OPTICAL Z (D-OP_PATH_LENGTH_DIFF+M3_Z_OFFSET) -(RASIUS_CURVATURE) ELLIPSE 2@(DIA/2) OBJECT '3RD_MIRROR' INTERFACE COATING REFLECT AIR BK7

GROUP -1

136 SHIFT Y (H1+H2+AOM_OFFSET) ROTATE X -(MIRROR_ROTATION_ANGLE) (H1+H2+AOM_OFFSET) (D- OP_PATH_LENGTH_DIFF+M3_Z_OFFSET) COLOR 3 !!END:3RD MIRROR

!!START: 4TH MIRROR SURFACE OPTICAL Z -(A+Z_OFFSET) 0 ELLIPSE 2@(DIA/2) OBJECT '4TH_MIRROR' INTERFACE COATING REFLECT AIR BK7

GROUP -1 ROTATE X -(MIRROR_ROTATION_ANGLE) 0 -(A+Z_OFFSET) COLOR 3 !!END:1ST MIRROR

!!START: DETECTOR SURFACE PLANE Z (DETECTOR_LOCATION) ELLIPSE 2@4 OBJECT 'DETECTOR' INTERFACE 0 0 AIR AIR

GROUP -1 SHIFT Y (H1+H2) !!ROTATE X (MIRROR_ROTATION_ANGLE) (H1+H2) (-D) COLOR 2 !!END: DETECTOR

MAKEGRAPH { GRAPH 'FIELD ENERGY ALONG OPTICAL AXIS FOR INITIAL WAIST=#1' 'POSITION IN MM' 'GRAPH MAXIMUM WILL BE NEW BEAM WAIST' 'ON-AXIS IRRADIACNE IS MAXIMUM AT WAIST' }

!! SOURCE DEFINITIONS BEGIN HERE

PARABASAL 4 BEAM COHERENT DIFFRACT WAVELENGTH 980 NM WIDTHS 1.6 GAUSSIAN Z (GAUSSIAN_ST_PT) 0 (NBR) (GAUSSIAN_BEAM_DIV_ANG) RADIANS FLUX 1 !! NORALIZE TOTAL FLUX !! TRACE RAYS WITH OPTIONAL PLOT

137

HALT 1

PLOT FACET 5 5 OVERLAY

FLUX 1 TRACE PLOT !!TRACE WINDOW X Y SPOTS POSITION

PIXELS 100 WINDOW Z (Z1) (Z2) Y [email protected] !! TRICK TO CALCULATE IRRADIACNCE ON AXIS

FIELD ENERGY 0 !!! DISPLAY !! PICTURE !!RETURN $MAKEGRAPH LIT(GAUSSIAN_BEAM_WAIST) !! !!DEPTH INFO RETURN

TRACE CONSIDER ONLY DETECTOR FOCUS MOVE

!!$GRAB 'Centroid Point' 0 1 X Y Z !!® WINDOW X [email protected] Y 173.38-0.05 173.38+0.05!!-2@3 PIXELS 100 SPREAD NORMAL

DISPLAY PICTURE '@Y=173.38 & Z=13' RETURN !!$VIEW

138 APPENDIX C: EXTERNAL CAVITY OF 2-SECTION SEMICONDUCTOR

MODE-LOCKED LASER

139 RF - DC

AR OC HR

Saturable Absorber

Fig. C - 1. 2-Section external cavity semiconductor mode-locked laser.

Passive mode-locking generates a shorter pulse than active mode-locking. A 2-section device with a saturable absorber is a good for a master oscillator. An external cavity with a 2- section device is demonstrated and investigated. Generally, modulation is imposed on the saturable absorber section to generate short pulse with a high repetition rate, such as 10GHz.

However, in this setup the gain section is modulated to accomplish a lower repetition rate mode- locking while the saturable absorber section is reversely biased as shown in Fig. C-1.

The performance of the mode-locked operation is summarized in Fig. C-2. A 519MHz of repetition rate, which is a cavity fundamental, at 53mA direct current is achieved. The average power of the external cavity is 0.76mW. A 1.55ps of auto-correlation width, which is compressed in a free space dispersion compensator, is achieved. The pulse width is ~ 1ps when assuming a hyperbolic secant pulse shape.

140 1.0 500ps/div 0.8

0.6

1.55 psec 0.4

A.C(a.u) SHG 0.2

0.0 Delay(3.1ps/div)

(a) (b)

1.0

0.8

0.6

0.4

0.2 Spectral Power(A.U) 0.0 965 970 975 980 Wavelength(nm)

(C)

Fig. C - 2. Performance of 2-Section external cavity semiconductor mode-locked laser.

141 APPENDIX D: NOVEL FREQUENCY SWEPT SOURCE OF MODE-

LOCKED SEMICONDUCTOR LASER

142 L red

L blue

SOA CFBG HR

Chirped RF Signal

Fig. D - 1. Proposed frequency swept semiconductor mode-locked laser.

Fig D-1 depicts a novel frequency swept source. The setup takes advantage of wide bandwidth a long chirped fiber bragg grating for a mode-locked semiconductor laser. A fixed repetition rate of the mode-locked laser limits band width of mode-locked operation in the experiment. If we match the swept rate of the chirped RF signal with the optical spectrum, the mode-locking is accomplished. In above picture, we used the residual transmittance of the CFBG as an output coupler.

One of features of the cavity is that the repetition rates vary with respect to the mode- locked band. This feature can be avoided when another chirped fiber bragg grating with an opposite group delay is used. If the pulses from the oscillator is circulated with the chirped fiber bragg grating, the repetition rate can be fixed with respect to the mode-locked spectrum. This scheme also provides additional pulse compression.

143 Another feature of the laser is that when the cavity is mode-locked with a harmonics of the cavity fundamental, multi-wave length operation can be achieved due to multiple resonances of the optical band in terms of repetition rate.

155.6MHz 38.9MHz 51.9MHz -20

-40 Spectral Power(dBm)

-60 970 972 974 976 978 980 Wavelength(nm)

Fig. D - 2. Mode-locked spectrum with fixed repetition rates.

Fig. D-2 shows mode-locked spectra of the linear cavity with respect to the repetition rate.

In this experiment, an 1800ps/nm of group delay and 7nm bandwidth centered at 975nm of the chirped fiber bragg grating is used. One can confirm the group delay of chirped fiber bragg grating from the mode-locked spectra.

This frequency swept semiconductor mode-locked laser is a good candidate of the light source for the high speed optical coherence tomography if bandwidth of chirped fiber bragg grating is sufficiently broad.

144 LIST OF REFERENCES

CHAPTER 1

[1] Hall, Robert N.; G. E. Fenner, J. D. Kingsley, T. J. Soltys, and R. O. Carlson, "Coherent

Light Emission From GaAs Junctions". Phys. Rev. Lett., 9, 9, 366 (1962).

[2] S. Gee, G. Alphonse, J. Connolly, and P. Delfyett, “High-power mode-locked external

cavity semiconductor laser using inverse bow-tie semiconductor optical amplifiers,” IEEE

J. Select. Topics Quant. Electron., 4, 209(1998).

[3] C. Pedersen and R. Hansen, "Single frequency, high power, tapered diode laser using

phase-conjugated feedback," Opt. Express 13, 3961 (2005).

[4] S. O'Brien, A. Schoenfelder, R.J. Lang, “5-W CW diffraction-limited InGaAs broad-area

flared amplifier at 970 nm,” IEEE Photon. Technol. Lett., 9, 9, 1217 (1997).

[5] P. Juodawlkis, 3. J. Plant, R IC Huang, L. J. Missaggia and J. P. Donnelly, “High-Power

1.5-um InGaAsP-InF Slab-Coupled Optical Waveguide Amplifier”, IEEE Photon. Technol.

Lett., 17, 2, 279 (2005).

[6] Zory, P. and Comerford, L. D “Grating-coupled double-heterostructure AlGaAs diode

lasers” IEEE J. Quantum Electron. 11, 451 (1975).

[7] Iga, Kenichi, "Surface-emitting laser—its birth and generation of new optoelectronics

field", IEEE J. Sel. Top. Quantum Electron., 6, 6, 1201 (2000).

[8] http://www.m2k-laser.de.

[9] Micheal J. Connelly, “Semiconductor Optical Amplifiers,” Kluwer Academic Publishers

(2002).

145 [10] M. Reed, T. M. Benson, P. C. Kendall, and P. Sewell, “Antireflection-coated angled facet

design,” in Inst. Elect. Eng. Proc.—Optoelectron., 143, 4, 214 (1996).

[11] J. Landreau and H. Nakajima, “In situ reflectivity monitoring of antireflection coatings on

semiconductor laser facets through facet loss induced forward voltage changes,” Appl.

Phys. Lett., 56, 24, 2376 (1990).

[12] G. Eisenstein, R. Tucker, U. Koren, S. Korotky, “Active mode-locking characteristics of

InGaAsP-single mode fiber composite cavity lasers,” IEEE J. Quantum Electron., 22,142

(1986).

[13] K. L. Hall, G. Lenz, E. P. Ippen, U. Koren, and G. Reybon, “Carrier heating and spectral

hole burning in strained-layer quantum-well laser amplifiers at 1.5 µm,” Appl. Phys. Lett.

61, 2512 (1992).

[14] J. Mork, M. L. Nielsen, and T. W. Berg, "The Dynamics of Semiconductor Optical

Amplifiers: Modeling and Applications," Optics & Photonics News 14, 42 (2003)

[15] P. J. Delfyett, L. T. Florez, N. Stoffel, T. Gmitter, N. C. Andreadakis, Y. Silberberg, J. P.

Heritage, and G. A. Alphonse, “High power ultrafast laser diodes,” IEEE J. Quantum

Electron., 28, 2203 (1992).

[16] G. P. Agrawal and N. A. Olsson, “Self phase modulation and spectral broadening of optical

pulses in semiconductor laser amplifiers,” IEEE J. Quantum Electron., 25, 2297 (1989).

[17] Peter Vasil’ev, “ Ultrafast Diode Lasers- Fundamentals and Applications,” Artech House

(1995)

146 CHAPTER 2

[1] S. Lee, K. Kim, Vaissie, L., Smolski, O.V., Johnson, E.G., and Delfyett, P.J., Jr.,

“Picosecond pulse generation using a saturable absorber section of grating-coupled surface-

emitting laser,” IEEE Photon. Technol. Lett., 17, 2676 (2005).

[2] R. Häring, R. Paschotta, A. Aschwanden, E. Gini, F. Morier-Genoud, and U. Keller,

“High-power Passively mode-locked semiconductor lasers,” IEEE J. Quantum Electron.,

38, 1268 (2002).

[3] A. Mar, R. Helkey, J. Bowers, D. Mehuys, and D. Welch, “Modelocked operation of a

master oscillator power amplifier,” IEEE Photon. Technol. Lett., 6, 1067 (1994).

[4] S. Gee, G. Alphonse, J. Connolly, and P. J. Delfyett, “High-power mode-locked external

cavity semiconductor laser using inverse bow-tie semiconductor optical amplifiers,” IEEE

J. Sel. Topic. Quantum Electron., 4, 209 (1998).

[5] K. Kim, S. Lee, and P. J. Delfyett, “1.4 kW high peak power generation from an all

semiconductor mode-locked oscillator power amplifier system based on eXtreme Chirped

Pulse Amplification (XCPA),” Opt. Express, 13, 4600 (2005).

[6] D. Strickland and G. Morou, “Compression of amplified chirped optical pulses,” Optics

Comm. 56, 219 (1985).

[7] Joseph T.Verdeyen, “Laser Electroncis,” 3rd Edition. 384 (2000).

[8] C. G. Gilbreath, W. S. Rabinovich, T. J. Meehan, M. J. Vilcheck, M. F. Stell, R. Mahon, P.

G. Goetz, E. Oh, J. A. Vasquez, K. Cochrell, R. Lucke, and S. Mozersky, "Progress in

development of multiple-quantum-well retromodulators for free-space data links," Opt.

Eng. 42 , 6, 1611 (2003).

147 [9] R. L. Fork, B. I. Greene, and C. V. Shank, “Generation of optical pulses shorter than 0.1

psec by colliding pulse mode locking,” Appl. Phys. Lett., 38, 671 (1981).

[10] B. S. Kim, Y. Chung, S. Lee, S. H. Kim, K. N. Kang, and S. S. Choi, “THz-class short-

pulse generation in mode-locked laser diodes with periodically sampled gratings,” in Eur.

Conf. Optical Communications ECOC '98, 479 (1998).

[11] M. C. Nuss, R. Leonhardt, and W. Zinth, “Stable operation of a synchronously pumped

colliding-pulse mode-locked ring dye laser,” Opt. Lett. 10, 16 (1985).

[12] A. E. Siegman, "Passive mode locking using an antiresonant-ring laser cavity," Opt. Lett.,

6, 334 (1981).

[13] E. B. Tracy, “Optical pulse compression with diffraction gratings,” IEEE J. Quantum

Electron., 5, 454 (1969).

[14] J. Brennan and D. LaBrake, “Realization of >10-m-long chirped fiber Bragg gratings,”

Bragg Gratings, Photosensitivity, and Poling in Glass Waveguides (OSA, Washington. DC,

1999), 128-130.

CHAPTER 3

[1] Bojan Resan, Luis Archundia, and Peter J. Delfyett, “Dispersion-managed semiconductor

mode-locked ring laser,” Opt. Lett. 28, 1371 (2003).

[2] Bojan Resan, and Peter J. Delfyett, Jr., “Dispersion-Managed Breathing-Mode

Semiconductor Mode-Locked Ring laser: Experimental Characterization and Numerical

Simulations,” IEEE J. Quant. Electron, 40, 214 (2004).

148 [3] J. F. Brennan and D. L. LaBrake, “Digest of Bragg Gratings, Photosensitivity, and Poling

in Glass Waveguide,” Optical Society of America, Washington, D. C., 128 (1999).

[4] G. P. Agrawal and N. A. Olsson, “Self-phase modulation and spectral broadening of optical

pulses in semiconductor laser amplifiers,” IEEE J. Quant. Electron. 25, 2297 (1989).

[5] P. J. Delfyett, L. T. Florez, N. Stoffel, T. Gmitter, N. C. Andreadakis, Y. Silberberg, J. P.

Heritage, and G. A. Alphonse, “High-Power Ultrafast Laser Diodes,” IEEE J. Quant.

Electron. 28. 2203 (1992).

[6] A. Mar, R. Helkey, J. Bowers, D. Mehuys, and D. Welch, “Modelocked operation of a

master oscillator power amplifier,” IEEE Photon. Technol. Lett., 6, 1067 (1994).

[7] A. S. Bhushan, F. Coppinger, B. Jalali, S. Wang, and H. F. Fetterman, “150 Gsample/s

wavelength division sampler with time-stretched output,” Electron. Lett, 34, 474 (1998).

[8] K. Kim, S. Lee, and P. J. Delfyett, “1.4kW high peak power generation from an all

semiconductor mode-locked master oscillator power amplifier system based on eXreme

Chirped Pulse amplification (XCPA)”,Opt. Express, 13, 4600 (2005),

[9] S. Lee, K. Kim, and P. J. Delfyett, “Extreme Chirped Pulse Oscillator (XCPO) Using a

Theta Cavity Design,” IEEE Photon. Technol. Lett., 8, 799 (2006).

[10] R. Huber, M. Wojtkowski, and J. G. Fujimoto, "Fourier Domain Mode Locking (FDML):

A new laser operating regime and applications for optical coherence tomography," Opt.

Express 14, 3225 (2006).

[11] F. Coppinger, A. S. Bhushan, and B. Jalali, “Photonic time stretch and its application to

analog-to-digital conversion,” IEEE Trans. Microwave Theory Tech., 47, 1309 (1999).

149 [12] P.W. Juodawlkis, J.J. Plant, R.K. Huang, L.J. Missaggia, J.P. Donnelly, “High-Power 1.5-

μm InGaAsP-InP Slab-Coupled Optical Waveguide Amplifier,” IEEE Photon. Technol.

Lett., 17, 279 (2005).

[13] S. Gee, F. Quinlan, S. Ozharar, P. J. Delfyett, J. J. Plant, and P. W. Juodawlkis, "Ultralow-

noise mode-locked optical pulse trains from an external cavity laser based on a slab

coupled optical waveguide amplifier (SCOWA)," Opt. Lett., 30, 2742 (2005).

CHAPTER 4

[1] Jiang, J., Smolski, O., Roychoudhuri, C., Portnoi, E., Venus, G., Gadjiev, I., McKillop,

J.,”Broad tunability of grating coupled surface-emitting laser with external cavity,”

Electron. Lett., 35, 1847 (1999).

[2] Uemukai, M., Matsumoto, N., Suhara, T., Nishihara, H., Eriksson, N., Larsson,

A.,”Monolithically integrated InGaAs-AlGaAs master oscillator power amplifier with

grating outcoupler,” IEEE Photon. Technol. Lett., 10, 1097 (1998).

[3] Mielke, M., Alphonse, G.A., Delfyett, P.J., “168 channels x 6 GHz from a multiwavelength

mode-locked semiconductor laser,” IEEE Photon. Technol. Lett., 15, 501 (2003).

[4] L. Vaissié, O. V. Smolski, and E. G. Johnson, "Crossed-beam superluminescent diode,"

Opt. Lett. 30, 1608 (2005).

[5] K. Kim, S. Lee, and P. Delfyett, "1.4kW high peak power generation from an all

semiconductor mode-locked master oscillator power amplifier system based on eXtreme

Chirped Pulse Amplification(X-CPA)," Opt. Express 13, 4600 (2005)

150 [6] Peter P. Vasil’ev, Valentin N. Morozov, Yuri M. Popov, Alexander B.

Sergeev.:’Subpicosecond pulse generation by a tandem-type AlGaAs DH laser with

colliding pulse mode locking’, IEEE J. Quantum Electron., 22, 149 (1986).

[7] K. Kim, S. Lee, O. Smolski, P. J. Delfyett, Jr.:’External-cavity, actively mode-locked

grating-coupled surface-emitting laser and amplification characteristics of a grating-

coupled semiconductor optical amplifier’, Opt. Lett., 29, 1273 (2004).

[8] K. Oe, Y. Noguchi, and C. Caneau, “GaInAsP lateral current injection lasers on semi-

insulating substrates,” IEEE Photon. Technol. Lett., 6, 479 (1994).

[9] D. J. Derickson, R. J. Helkey, A. Mar, J. R. Karin, J. G. Wasserbauer, and J. E. Bowers,

“Short pulse generation using multisegment mode-locked semiconductor lasers,” IEEE J.

Quantum Electron., 28, 2186 (1992).

[10] S. Arahira, Y. Matsui, and Y. Ogawa, “Mode-locking at very high repetition rates more

than terahertz in passively mode-locked distributed-Bragg-reflector laser diodes,” IEEE J.

Quantum Electron., 32, 1211 (1996).

[11] Fukuda, M. Reliability and Degradation of Semiconductor Lasers and LEDs, Artech,

Boston (1991).

[12] L. J. Mawst, A. Bhattacharya, J. Lopez, D. Botez, D. Z. Garbuzov, L. DeMarco, J. C.

Connolly, M. Jansen, F. Fang, and R. F. Nabiev, “8 W continuous wave front-facet power

from broad-waveguide Al-free 980 nm diode lasers,” Appl. Phys. Lett., 69, 11, 1532 (1996).

[13] Erbert, G. Beister, G. Hulsewede, R. Knauer, A. Pittroff, W. Sebastian, J. Wenzel,

H. Weyers, M. Trankle, G.,”High-Power Highly Reliable Al-Free 940-nm Diode

Lasers,” IEEE, J. Sel. Topic. Quantum Electron., 7, 2, 143 (2001).

151 [14] A. Al-Muhanna, L.J. Mawst, D. Botez, D.Z. Garbuzov, R.U. Martinelli and J.C.

Conolly, ”High-power (>10 W) continuous-wave operation from 100-µm-aperture 0.97-

µm-emitting Al-free diode lasers,” Appl. Phys. Lett. 73, 1182 (1998).

[15] IGA, K., “Surface emitting lasers and parallel operating devices - fundamentals and

prospects,” IEICE Trans., E75-A, 12 (1992).

CHAPTER 5

[1] R.J. Jones and J. Ye, “Femtosecond pulse amplification by coherent addition in a passive

optical cavity,” Opt. Lett. 27, 1848 (2002).

[2] A. E. Siegman, “Laser,” Oxford University Press, 421-426 (1986).

[3] T. W. Hansch and B. Couillaud, "Laser frequency stabilization by polarization

spectroscopy of a reflecting reference cavity," Opt. Comm. 35, 441 (1980).

[4] R. W. P. Drever, J. L. Hall, F. V. Kowalski, J. Hough, G. M. Ford, A. J. Munley and H.

Ward, "Laser phase and frequency stabilization using an optical resonator," Appl. Phys. B,

31, 97 (1983).

[5] L. S. Cruz and F. C. Cruz, "External power-enhancement cavity versus intracavity

frequency doubling of Ti:sapphire lasers using BIBO," Opt. Express 15, 11913 (2007)

[6] S. Gee, F. Quinlan, S. Ozharar and P. J. Delfyett, "Simultaneous optical comb frequency

stabilization and super mode noise suppression of harmonically modelocked semiconductor

ring laser using an intracavity etalon," IEEE Photon. Technol. Lett., 17, 1, 199 (2005).

[7] E. O. Potma, C. Evans, X. S. Xie, R. J. Jones, and J. Ye, "Picosecond-pulse amplification

with an external passive optical cavity," Opt. Lett. 28, 1835 (2003).

152 [8] R. J. Jones and J. Ye, "High-repetition-rate coherent femtosecond pulse amplification with

an external passive optical cavity," Opt. Lett. 29, 2812 (2004).

[9] C. Gohle, T. Udem, M. Hermann, J. Rauschenberger, R. Holzwarth, H. A. Schuessler, F.

Krausz, and T. W. Hänsch, “A frequency comb in the extreme ultraviolet,” Nature 436, 234

(2005).

[10] http://www.toptica.com/products/itemlayer/110/BR-029-ProductCatalog-2007-06-

FCDL.pdf.

[11] H. Kogelnik, E. P. Ippen, A. Dienes, and C. V. Shank, “Astigmatically compensated

cavities for CW dye lasers,” IEEE J. Quantum. Electron. 8, 373 (1972).

[12] Kyungbum Kim, “ALL-SEMICONDUCTOR HIGH POWER MODE-LOCKED LASER

SYSTEM,” Ph.D Dissertation 104-105, University of Central Florida (2006).

CHAPTER 6

[1] K. Kim, S. Lee, and P. Delfyett, "1.4kW high peak power generation from an all

semiconductor mode-locked master oscillator power amplifier system based on eXtreme

Chirped Pulse Amplification(X-CPA)," Opt. Express 13, 4600 (2005).

[2] S. Lee, K. Kim, and P. J. Delfyett, “Extreme Chirped Pulse Oscillator (XCPO) Using a

Theta Cavity Design,” IEEE Photon. Technol. Lett., 8, 799 (2006).

[3] Shinwook Lee, Dimitrios Mandridis, and Peter J. Delfyett, Jr, “Extreme Chirped Pulse

Oscillator,” Opt.Express (submitted).

153 [4] S. Lee, K. Kim, Vaissie, L., Smolski, O.V., Johnson, E.G., and Delfyett, P.J., Jr.,

“Picosecond pulse generation using a saturable absorber section of grating-coupled surface-

emitting laser,” IEEE Photon. Technol. Lett., 17, 2676 (2005).

[5] Jones, R. J., Moll, K., Thorpe, M., Ye, J., “Phase-coherent frequency combs in the vacuum

ultraviolet via high-harmonic generation inside a femtosecond enhancement cavity,” Phys.

Rev. Lett. 94, 193201 (2005)

[6] Christoph Gohle, Thomas Udem, Maximilian Herrmann, Jens Rauschenberger, Ronald

Holzwarth, Hans A. Schuessler, Ferenc Krausz and Theodor W. Hänsch, “A frequency

comb in the extreme ultraviolet,” Nature 436, 234 (2005).

[7] G. C. Valley, "Photonic analog-to-digital converters," Opt. Express 15, 1955 (2007).

[8] H. F. Taylor, H. F., M. J. Taylor, and P. W. Bauer, "Electro-optic analog-to-digital

conversion using channel waveguide modulators," Appl. Phys. Lett. 32, 559 (1978).

[9] R. H. Walden, "Analog-to-digital converter survey and analysis," IEEE J. Sel. Areas

Comm., 17, 539 (1999).

[10] Delfyett PJ, DePriest C, Yilmaz, “Signal Processing at the Speed of Lightwaves,” IEEE

CIRCUITS & DEVICES 18 (5): 28-35 SEP 2002.

[11] A. S. Bhushan, F. Coppinger, B. Jalali, S. Wang and H. F. Fetterman, "150 Gsample/s

wavelength division sampler with time-stretched output," Electron. Lett., 34, 474 (1998).

[12] R. Huber, M. Wojtkowski, and J. G. Fujimoto, "Fourier Domain Mode Locking (FDML):

A new laser operating regime and applications for optical coherence tomography," Opt.

Express 14, 3225 (2006).

[13] G. Imeshev, I. Hartl, and M. E. Fermann, "Chirped pulse amplification with a nonlinearly

chirped fiber Bragg grating matched to the Treacy compressor," Opt. Lett. 29, 679 (2004).

154